Skip to contentSkip to conversation

Qemu: Enable dynamic hostcache change through monitor

PATCH Superseded Supriya Kannery 5 participants 11 messages May 19, 2011 – May 24, 2011

v2 of 9 Superseded Latest is v9 →

lore
Share & tools
lore
#0 lore raw

v2 cover Qemu: Enable dynamic hostcache change through monitor

 Currently host page cache setting for a block device cannot be changed without restarting a running VM. Following patchset [V2] is for enabling dynamic change of host cache setting for devices through qemu monitor.

Changes from patchset V1:
1. Support of dynamic cache change only for hostcache.
2. Monitor command "hostcache_get" added to display current cache setting
3. Backed off the changes for display of cache setting in "info block"

Two new monitor commands added.
1. hostcache_get
-- Displays hostcache setting of a block device.
Usage:
"hostcache_get <device>
<device> = block device

2. hostcache_set
-- Sets hostcache for the device to specified value.
Usage:
"hostcache_set <device> <cache>"
<device> = block device
<cache> = "on"/"off"

1/2 New error classes for file reopen and device insertion 2/2 Add commands "hostcache_set" and "hostcache_get"

 block.c         |   48 ++++++++++++++++++++++++++++++++++++++++++++++++
 block.h         |    2 ++
 blockdev.c      |   48 ++++++++++++++++++++++++++++++++++++++++++++++++
 blockdev.h      |    2 ++
 hmp-commands.hx |   29 +++++++++++++++++++++++++++++
 qerror.c        |    8 ++++++++
 qerror.h        |    6 ++++++
 qmp-commands.hx |   55 
 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 8 files changed, 198 insertions(+)

~
~
~

#1 lore raw

PATCH v2 1/2 Qemu: New error classes for file reopen and device insertion

New errors defined for device insertion and file reopen

---
 qerror.c |    8 ++++++++
 qerror.h |    6 ++++++
 2 files changed, 14 insertions(+)
qerror.c +8 -0
Index: qemu/qerror.c
===================================================================
@@ -93,6 +93,10 @@ static const QErrorStringTable qerror_ta
         .desc      = "Device '%(device)' not found",
     },
     {
+        .error_fmt = QERR_DEVICE_NOT_INSERTED,
+        .desc      = "Device '%(device)' has not been inserted",
+    },
+    {
         .error_fmt = QERR_DEVICE_NOT_REMOVABLE,
         .desc      = "Device '%(device)' is not removable",
     },
@@ -161,6 +165,10 @@ static const QErrorStringTable qerror_ta
         .desc      = "Could not open '%(filename)'",
     },
     {
+        .error_fmt = QERR_REOPEN_FILE_FAILED,
+        .desc      = "Could not reopen '%(filename)'",
+    },
+    {
         .error_fmt = QERR_PROPERTY_NOT_FOUND,
         .desc      = "Property '%(device).%(property)' not found",
     },
qerror.h +6 -0
Index: qemu/qerror.h
===================================================================
@@ -84,6 +84,9 @@ QError *qobject_to_qerror(const QObject 
 #define QERR_DEVICE_NOT_FOUND \
     "{ 'class': 'DeviceNotFound', 'data': { 'device': %s } }"
 
+#define QERR_DEVICE_NOT_INSERTED \
+    "{ 'class': 'DeviceNotInserted', 'data': { 'device': %s } }"
+
 #define QERR_DEVICE_NOT_REMOVABLE \
     "{ 'class': 'DeviceNotRemovable', 'data': { 'device': %s } }"
 
@@ -135,6 +138,9 @@ QError *qobject_to_qerror(const QObject 
 #define QERR_OPEN_FILE_FAILED \
     "{ 'class': 'OpenFileFailed', 'data': { 'filename': %s } }"
 
+#define QERR_REOPEN_FILE_FAILED \
+    "{ 'class': 'ReopenFileFailed', 'data': { 'filename': %s } }"
+
 #define QERR_PROPERTY_NOT_FOUND \
     "{ 'class': 'PropertyNotFound', 'data': { 'device': %s, 'property': %s } }"
 
#2 lore raw

PATCH v2 2/2 Qemu: Add commands "hostcache_set" and "hostcache_get"

In reply to Supriya Kannery

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

---
 block.c         |   48 ++++++++++++++++++++++++++++++++++++++++++++++++
 block.h         |    2 ++
 blockdev.c      |   48 ++++++++++++++++++++++++++++++++++++++++++++++++
 blockdev.h      |    2 ++
 hmp-commands.hx |   29 +++++++++++++++++++++++++++++
 qmp-commands.hx |   55 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 6 files changed, 184 insertions(+)
hmp-commands.hx +29 -0
Index: qemu/hmp-commands.hx
===================================================================
@@ -70,6 +70,35 @@ but should be used with extreme caution.
 resizes image files, it can not resize block devices like LVM volumes.
 ETEXI
 
+    {
+        .name       = "hostcache_get",
+        .args_type  = "device:B",
+        .params     = "device",
+        .help       = "retrieve host cache settings for device",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_hostcache_get,
+    },
+
+STEXI
+@item hostcache_get
+@findex hostcache_get
+Display host cache settings for a block device while guest is running.
+ETEXI
+
+    {
+        .name       = "hostcache_set",
+        .args_type  = "device:B,hostcache:s",
+        .params     = "device hostcache",
+        .help       = "change host cache setting for device",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_hostcache_set,
+    },
+
+STEXI
+@item hostcache_set
+@findex hostcache_set
+Change host cache options for a block device while guest is running.
+ETEXI
 
     {
         .name       = "eject",
block.c +48 -0

57 lines not on this page · raw message

blockdev.c +48 -0

51 lines not on this page · raw message

block.h +2 -0
Index: qemu/block.h
===================================================================
@@ -71,6 +71,7 @@ void bdrv_delete(BlockDriverState *bs);
 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags);
 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
               BlockDriver *drv);
+int bdrv_reopen(BlockDriverState *bs, int bdrv_flags);
 void bdrv_close(BlockDriverState *bs);
 int bdrv_attach(BlockDriverState *bs, DeviceState *qdev);
 void bdrv_detach(BlockDriverState *bs, DeviceState *qdev);
@@ -96,6 +97,7 @@ void bdrv_commit_all(void);
 int bdrv_change_backing_file(BlockDriverState *bs,
     const char *backing_file, const char *backing_fmt);
 void bdrv_register(BlockDriver *bdrv);
+int bdrv_change_hostcache(BlockDriverState *bs, bool enable_host_cache);
 
 
 typedef struct BdrvCheckResult {
blockdev.h +2 -0
Index: qemu/blockdev.h
===================================================================
@@ -64,5 +64,7 @@ int do_change_block(Monitor *mon, const 
 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);
+int do_hostcache_get(Monitor *mon, const QDict *qdict, QObject **ret_data);
+int do_hostcache_set(Monitor *mon, const QDict *qdict, QObject **ret_data);
 
 #endif
qmp-commands.hx +55 -0

61 lines not on this page · raw message

#3 lore raw

Re: [Qemu-devel] [V2 2/2]Qemu: Add commands "hostcache_set" and "hostcache_get"

Quoted 2 lines · Supriya Kannery

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

A generic command for changing block device options would be nice, althought I don't see other options where it makes sense to change them at runtime.

The alternative would be:

block_set hostcache on

"block_set", {"device": "ide1-cd0", "name": "hostcache", "enable": true}

The hostcache_get information would be part of query-block output:

         {
            "device":"ide0-hd0",
            "locked":false,
            "removable":false,
            "inserted":{
               "ro":false,
               "drv":"qcow2",
               "encrypted":false,
               "file":"disks/test.img"
	       "hostcache":true,
            },
            "type":"hd"
         },

This approach is extensible if more options need to be exposed.

Quoted 16 lines
---
 block.c         |   48 ++++++++++++++++++++++++++++++++++++++++++++++++
 block.h         |    2 ++
 blockdev.c      |   48 ++++++++++++++++++++++++++++++++++++++++++++++++
 blockdev.h      |    2 ++
 hmp-commands.hx |   29 +++++++++++++++++++++++++++++
 qmp-commands.hx |   55 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 6 files changed, 184 insertions(+)

Index: qemu/hmp-commands.hx
===================================================================

hmp-commands.hx but should be used with extreme caution. · line 77

Quoted 8 lines
hmp-commands.hx +5 -0
@@ -70,6 +70,35 @@ but should be used with extreme caution.
 resizes image files, it can not resize block devices like LVM volumes.
 ETEXI
 
+    {
+        .name       = "hostcache_get",
+        .args_type  = "device:B",
+        .params     = "device",
+        .help       = "retrieve host cache settings for device",

Please make it clear these operations affect block devices:
"for block device"

Quoted 48 lines
hmp-commands.hx +24 -0
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_hostcache_get,
+    },
+
+STEXI
+@item hostcache_get
+@findex hostcache_get
+Display host cache settings for a block device while guest is running.
+ETEXI
+
+    {
+        .name       = "hostcache_set",
+        .args_type  = "device:B,hostcache:s",
+        .params     = "device hostcache",
+        .help       = "change host cache setting for device",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_hostcache_set,
+    },
+
+STEXI
+@item hostcache_set
+@findex hostcache_set
+Change host cache options for a block device while guest is running.
+ETEXI
 
     {
         .name       = "eject",

Index: qemu/block.c
===================================================================

block.c +13 -0
@@ -657,6 +657,34 @@ unlink_and_fail:
     return ret;
 }
 
+int bdrv_reopen(BlockDriverState *bs, int bdrv_flags)
+{
+    BlockDriver *drv = bs->drv;
+    int ret = 0;
+
+    /* No need to reopen as no change in flags */
+    if (bdrv_flags == bs->open_flags) {
+        return 0;
+    }
+
+    /* Quiesce IO for the given block device */
+    qemu_aio_flush();
+    bdrv_flush(bs);

block.c unlink_and_fail: · line 682

Quoted 10 lines
block.c +10 -0
+
+    bdrv_close(bs);
+    ret = bdrv_open(bs, bs->filename, bdrv_flags, drv);
+
+    /*
+     * A failed attempt to reopen the image file must lead to 'abort()'
+     */
+    if (ret != 0) {
+        qerror_report(QERR_REOPEN_FILE_FAILED, bs->filename);
+        abort();

The error is never reported on a QMP monitor because qerror_report() simply stashes away the qerror. The QMP client doesn't have a chance to read the error before QEMU terminates.

Quoted 9 lines
block.c +5 -0
+    }
+
+    return ret;
+}
+
 void bdrv_close(BlockDriverState *bs)
 {
     if (bs->drv) {

block.c out: · line 3081

Quoted 5 lines
block.c +2 -0
@@ -3049,3 +3077,23 @@ out:
 
     return ret;
 }
+
+int bdrv_change_hostcache(BlockDriverState *bs, bool enable_host_cache)

Consistently using "hostcache" or "host_cache" would be nice.

block.c out: · line 3086

Quoted 5 lines
block.c +5 -0
+{
+    int bdrv_flags = bs->open_flags;
+
+    /* No change in existing hostcache setting */
+    if(!enable_host_cache == (bdrv_flags & BDRV_O_NOCACHE)) {

This expression doesn't work as expected. bool has a lower rank than int. That means !enable_host_cache is converted to an int and compared against bdrv_flags & BDRV_O_NOCACHE. This expression is always false because a bool is 0 or 1 and BDRV_O_NOCACHE is 0x0020.

block.c out: · line 3087

Quoted 1 line
block.c +1 -0
+        return -1;

This shouldn't be a failure and please don't use -1 when a negative errno indicates failure. -1 == -EPERM. The return value should be 0 here.

block.c out: · line 3088

Quoted 1 line
block.c +1 -0
+    }

Anyway, this whole check is unnecessary since bdrv_reopen() already performs it.

block.c out: · line 3098

Quoted 10 lines
block.c +10 -0
+
+    /* set hostcache flags (without changing WCE/flush bits) */
+    if(!enable_host_cache) {
+        bdrv_flags |= BDRV_O_NOCACHE;
+    } else {
+        bdrv_flags &= ~BDRV_O_NOCACHE;
+    }
+
+    /* Reopen file with changed set of flags */
+    return(bdrv_reopen(bs, bdrv_flags));

Please run scripts/checkpatch.pl before submitting patches.

Quoted 41 lines
block.c +1 -0
+}

Index: qemu/blockdev.c
===================================================================

blockdev.c +32 -0
@@ -796,3 +796,51 @@ int do_block_resize(Monitor *mon, const 
 
     return 0;
 }
+
+int do_hostcache_get(Monitor *mon, const QDict *qdict, QObject **ret_data)
+{
+    const char *device = qdict_get_str(qdict, "device");
+    BlockDriverState *bs;
+
+    bs = bdrv_find(device);
+    if (!bs) {
+        qerror_report(QERR_DEVICE_NOT_FOUND, device);
+        return -1;
+    }
+
+    monitor_printf(mon, "%s:", device);
+
+    if (bs->open_flags & BDRV_O_NOCACHE) {
+        monitor_printf(mon, " hostcache=off\n");
+    } else {
+        monitor_printf(mon, " hostcache=on\n");
+    }
+
+    return 0;
+}
+
+int do_hostcache_set(Monitor *mon, const QDict *qdict, QObject **ret_data)
+{
+    const char *device = qdict_get_str(qdict, "device");
+    const char *hostcache = qdict_get_str(qdict, "hostcache");
+    BlockDriverState *bs;
+    bool enable_host_cache = 0;
+
+    bs = bdrv_find(device);
+    if (!bs) {

blockdev.c int do_block_resize(Monitor *mon, const · line 840

Quoted 10 lines
blockdev.c +10 -0
+        qerror_report(QERR_DEVICE_NOT_FOUND, device);
+        return -1;
+    }
+
+    /* cache change applicable only if device inserted */
+    if (bdrv_is_inserted(bs)) {
+        if (!strcmp(hostcache,"on"))
+            enable_host_cache = 1;
+        else
+            enable_host_cache = 0;

Coding style, please use {} or remove the if:
enable_host_cache = !strcmp(hostcache, "on");

Quoted 49 lines
blockdev.c +6 -0
+        return(bdrv_change_hostcache(bs, enable_host_cache));
+    } else {
+       qerror_report(QERR_DEVICE_NOT_INSERTED, device);
+       return -1;
+    }
+}

Index: qemu/block.h
===================================================================

block.h +2 -0
@@ -71,6 +71,7 @@ void bdrv_delete(BlockDriverState *bs);
 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags);
 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
               BlockDriver *drv);
+int bdrv_reopen(BlockDriverState *bs, int bdrv_flags);
 void bdrv_close(BlockDriverState *bs);
 int bdrv_attach(BlockDriverState *bs, DeviceState *qdev);
 void bdrv_detach(BlockDriverState *bs, DeviceState *qdev);
@@ -96,6 +97,7 @@ void bdrv_commit_all(void);
 int bdrv_change_backing_file(BlockDriverState *bs,
     const char *backing_file, const char *backing_fmt);
 void bdrv_register(BlockDriver *bdrv);
+int bdrv_change_hostcache(BlockDriverState *bs, bool enable_host_cache);
 
 
 typedef struct BdrvCheckResult {

Index: qemu/blockdev.h
===================================================================

blockdev.h +2 -0
@@ -64,5 +64,7 @@ int do_change_block(Monitor *mon, const 
 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);
+int do_hostcache_get(Monitor *mon, const QDict *qdict, QObject **ret_data);
+int do_hostcache_set(Monitor *mon, const QDict *qdict, QObject **ret_data);
 
 #endif

Index: qemu/qmp-commands.hx
===================================================================

qmp-commands.hx +3 -0
@@ -664,6 +664,61 @@ Example:
 -> { "execute": "block_resize", "arguments": { "device": "scratch", "size": 1073741824 } }
 <- { "return": {} }
 
+
+EQMP
+

qmp-commands.hx Example: · line 679

Quoted 10 lines
qmp-commands.hx +10 -0
+    {
+        .name       = "hostcache_get",
+        .args_type  = "device:B",
+        .params     = "device",
+        .help       = "retrieve host cache settings for device",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_hostcache_get,
+    },
+SQMP
+cache_get

hostcache_get

qmp-commands.hx Example: · line 682

Quoted 3 lines
qmp-commands.hx +3 -0
+---------
+
+Retrieve host page cache setting while a guest is running.

Please make it clear that this is an operation on block devices.

Quoted 22 lines
qmp-commands.hx +22 -0
+
+Arguments:
+
+- "device": the device's ID, must be unique (json-string)
+
+Example:
+
+-> { "execute": "hostcache_set", "arguments": { "device": "ide0-hd0" } }
+<- { "return": {} }
+
+
+EQMP
+
+    {
+        .name       = "hostcache_set",
+        .args_type  = "device:B,cache:s",
+        .params     = "device cache",
+        .help       = "change hostcache setting for device",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_hostcache_set,
+    },
+

qmp-commands.hx Example: · line 714

Quoted 10 lines
qmp-commands.hx +10 -0
+SQMP
+hostcache_set
+-------------
+
+Change host page cache setting while a guest is running.
+
+Arguments:
+
+- "device": the device's ID, must be unique (json-string)
+- "cache": host cache value "off" or "on" (json-string)

There is a boolean value that can be used instead of string on|off. See the set_link command.

Stefan

#4 lore raw

Re: [Qemu-devel] [V2 2/2]Qemu: Add commands "hostcache_set" and "hostcache_get"

Quoted 29 lines · Stefan Hajnoczi
Quoted 2 lines · Supriya Kannery

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

A generic command for changing block device options would be nice, althought I don't see other options where it makes sense to change them at runtime.

The alternative would be:

block_set hostcache on

"block_set", {"device": "ide1-cd0", "name": "hostcache", "enable": true}

The hostcache_get information would be part of query-block output:

          {
             "device":"ide0-hd0",
             "locked":false,
             "removable":false,
             "inserted":{
                "ro":false,
                "drv":"qcow2",
                "encrypted":false,
                "file":"disks/test.img"
	       "hostcache":true,
             },
             "type":"hd"
          },
Quoted 1 line

This approach is extensible if more options need to be exposed.

Sure, I will resubmit this patchset, after making this feature more generic. Can you pls help finding atleast one or two options (other than hostcache) which can be changed dynamically. This will help me evaluate the generic approach.

Quoted 5 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+        .params     = "device",
+        .help       = "retrieve host cache settings for device",

Please make it clear these operations affect block devices:
"for block device"

ok

Quoted 8 lines

block.c unlink_and_fail: · line 682

Quoted 7 lines
block.c +6 -0
+    /*
+     * A failed attempt to reopen the image file must lead to 'abort()'
+     */
+    if (ret != 0) {
+        qerror_report(QERR_REOPEN_FILE_FAILED, bs->filename);
+        abort();
Quoted 4 lines

The error is never reported on a QMP monitor because qerror_report() simply stashes away the qerror. The QMP client doesn't have a chance to read the error before QEMU terminates.

Verified this from qemu command line and the error got displayed before aborting (when the image file was renamed while VM was running).

Quoted 5 lines

block.c out: · line 3081

Quoted 2 lines
block.c +2 -0
+
+int bdrv_change_hostcache(BlockDriverState *bs, bool enable_host_cache)

Consistently using "hostcache" or "host_cache" would be nice.

ok

Quoted 12 lines
Quoted 2 lines

+ return -1;

This shouldn't be a failure and please don't use -1 when a negative errno indicates failure. -1 == -EPERM. The return value should be 0 here.

Quoted 1 line

+ }

Anyway, this whole check is unnecessary since bdrv_reopen() already performs it.

will take this off

Quoted 2 lines

Please run scripts/checkpatch.pl before submitting patches.

ok..will do

Quoted 9 lines

qmp-commands.hx Example: · line 714

Quoted 5 lines
qmp-commands.hx +5 -0
+
+Arguments:
+
+- "device": the device's ID, must be unique (json-string)
+- "cache": host cache value "off" or "on" (json-string)

There is a boolean value that can be used instead of string on|off. See the set_link command.

ok

Quoted 2 lines

Stefan

#5 lore raw

Re: [Qemu-devel] [V2 2/2]Qemu: Add commands "hostcache_set" and "hostcache_get"

Quoted 32 lines · Supriya Kannery
Quoted 31 lines · Stefan Hajnoczi
Quoted 3 lines · Supriya Kannery

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

A generic command for changing block device options would be nice, althought I don't see other options where it makes sense to change them at runtime.

The alternative would be:

block_set hostcache on

"block_set", {"device": "ide1-cd0", "name": "hostcache", "enable": true}

The hostcache_get information would be part of query-block output:
         {
            "device":"ide0-hd0",
            "locked":false,
            "removable":false,
            "inserted":{
               "ro":false,
               "drv":"qcow2",
               "encrypted":false, "file":"disks/test.img" "hostcache":true,
            },
            "type":"hd"
         },

Quoted 6 lines
Quoted 1 line

This approach is extensible if more options need to be exposed.

Sure, I will resubmit this patchset, after making this feature more generic. Can you pls help finding atleast one or two options (other than hostcache) which can be changed dynamically. This will help me evaluate the generic approach.

Hang on, let's see if we can get agreement from Kevin and others before taking this approach. Like I said, I don't see other options that should be changed at runtime.

Quoted 17 lines
Quoted 6 lines
Quoted 3 lines

+        .params     = "device",
+        .help       = "retrieve host cache settings for device",

Please make it clear these operations affect block devices:
"for block device"

ok

Quoted 8 lines
Quoted 7 lines

+    /*
+     * A failed attempt to reopen the image file must lead to 'abort()'
+     */
+    if (ret != 0) {
+        qerror_report(QERR_REOPEN_FILE_FAILED, bs->filename);
+        abort();

Quoted 7 lines
Quoted 4 lines

The error is never reported on a QMP monitor because qerror_report() simply stashes away the qerror.  The QMP client doesn't have a chance to read the error before QEMU terminates.

Verified this from qemu command line and the error got displayed before aborting (when the image file was renamed while VM was running).

QMP takes a different code path from the human monitor (HMP) that you used; see monitor_cur_is_qmp() and how it affects the qerror_report() path. CCing Luiz to see whether there is a strategy that we should use here to ensure the error message does get delivered to the QMP client before exiting.

Stefan

#6 lore raw

Re: [Qemu-devel] [V2 2/2]Qemu: Add commands "hostcache_set" and "hostcache_get"

Am 23.05.2011 12:00, schrieb Stefan Hajnoczi:

Quoted 35 lines · Supriya Kannery
Quoted 34 lines · Stefan Hajnoczi
Quoted 32 lines · Supriya Kannery
Quoted 3 lines

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

A generic command for changing block device options would be nice, althought I don't see other options where it makes sense to change them at runtime.

The alternative would be:

block_set hostcache on

"block_set", {"device": "ide1-cd0", "name": "hostcache", "enable": true}

The hostcache_get information would be part of query-block output:

         {
            "device":"ide0-hd0",
            "locked":false,
            "removable":false,
            "inserted":{
               "ro":false,
               "drv":"qcow2",
               "encrypted":false,
               "file":"disks/test.img"
              "hostcache":true,
            },
            "type":"hd"
         },

This approach is extensible if more options need to be exposed.

Quoted 8 lines
Quoted 4 lines

Sure, I will resubmit this patchset, after making this feature more generic. Can you pls help finding atleast one or two options (other than hostcache) which can be changed dynamically. This will help me evaluate the generic approach.

Hang on, let's see if we can get agreement from Kevin and others before taking this approach. Like I said, I don't see other options that should be changed at runtime.

Things like enabling copy on read could fit here.

Generally I'm in favour of having a generic command. We just need to pay attention not to include things that we don't want to maintain long term, i.e. just putting the current cache=... parameter into the argument isn't going to work. Maybe two booleans 'o_direct' and 'ignore_flushes' is what we want to have. The same structure should be used for blkdev_add then, even though it will allow some more options.

I'm also not completely sure how you would enable cache=writethrough from the command line in a fully converted world. Would this be one of the arguments that must be specified on BlockDriverState creation and can't be changed later, and the device will pick it up from there? Or is it a qdev property and somehow makes it way to the block layer?

Kevin

#7 lore raw

Re: [Qemu-devel] [V2 2/2]Qemu: Add commands "hostcache_set" and "hostcache_get"

Quoted 45 lines · Kevin Wolf

Am 23.05.2011 12:00, schrieb Stefan Hajnoczi:

Quoted 35 lines · Supriya Kannery
Quoted 34 lines · Stefan Hajnoczi
Quoted 32 lines · Supriya Kannery
Quoted 3 lines

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

A generic command for changing block device options would be nice, althought I don't see other options where it makes sense to change them at runtime.

The alternative would be:

block_set hostcache on

"block_set", {"device": "ide1-cd0", "name": "hostcache", "enable": true}

The hostcache_get information would be part of query-block output:
         {
            "device":"ide0-hd0",
            "locked":false,
            "removable":false,
            "inserted":{
               "ro":false,
               "drv":"qcow2",
               "encrypted":false, "file":"disks/test.img" "hostcache":true,
            },
            "type":"hd"
         },

This approach is extensible if more options need to be exposed.

Quoted 8 lines
Quoted 4 lines

Sure, I will resubmit this patchset, after making this feature more generic. Can you pls help finding atleast one or two options (other than hostcache) which can be changed dynamically. This will help me evaluate the generic approach.

Hang on, let's see if we can get agreement from Kevin and others before taking this approach.  Like I said, I don't see other options that should be changed at runtime.

Quoted 8 lines

Things like enabling copy on read could fit here.

Generally I'm in favour of having a generic command. We just need to pay attention not to include things that we don't want to maintain long term, i.e. just putting the current cache=... parameter into the argument isn't going to work. Maybe two booleans 'o_direct' and 'ignore_flushes' is what we want to have. The same structure should be used for blkdev_add then, even though it will allow some more options.

Supriya, it sounds to me like the generic block_set command and query-block integration is an acceptable approach.

Stefan

#8 lore raw

Re: [Qemu-devel] [V2 2/2]Qemu: Add commands "hostcache_set" and "hostcache_get"

In reply to Kevin Wolf
Quoted 54 lines · Kevin Wolf

Am 23.05.2011 12:00, schrieb Stefan Hajnoczi:

Quoted 35 lines · Supriya Kannery
Quoted 34 lines · Stefan Hajnoczi
Quoted 32 lines · Supriya Kannery
Quoted 3 lines

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

A generic command for changing block device options would be nice, althought I don't see other options where it makes sense to change them at runtime.

The alternative would be:

block_set hostcache on

"block_set", {"device": "ide1-cd0", "name": "hostcache", "enable": true}

The hostcache_get information would be part of query-block output:

         {
            "device":"ide0-hd0",
            "locked":false,
            "removable":false,
            "inserted":{
               "ro":false,
               "drv":"qcow2",
               "encrypted":false,
               "file":"disks/test.img"
              "hostcache":true,
            },
            "type":"hd"
         },

This approach is extensible if more options need to be exposed.

Quoted 8 lines
Quoted 4 lines

Sure, I will resubmit this patchset, after making this feature more generic. Can you pls help finding atleast one or two options (other than hostcache) which can be changed dynamically. This will help me evaluate the generic approach.

Hang on, let's see if we can get agreement from Kevin and others before taking this approach. Like I said, I don't see other options that should be changed at runtime.

Things like enabling copy on read could fit here.

Generally I'm in favour of having a generic command. We just need to pay attention not to include things that we don't want to maintain long term, i.e. just putting the current cache=... parameter into the argument isn't going to work. Maybe two booleans 'o_direct' and 'ignore_flushes' is what we want to have. The same structure should be used for blkdev_add then, even though it will allow some more options.

Quoted 5 lines

I'm also not completely sure how you would enable cache=writethrough from the command line in a fully converted world. Would this be one of the arguments that must be specified on BlockDriverState creation and can't be changed later, and the device will pick it up from there? Or is it a qdev property and somehow makes it way to the block layer?

qdev properties are the configurable bits of the device's guest part. For something to be a qdev property, it must belong to the guest part, and it must be configurable.

Example: a drive serial number belongs to the guest part. It's guest-visible. The host part doesn't care about it; in fact, if you unplug the device, you could reuse the same host part with another guest part with different serial number.

Example: the image format doesn't belong to the guest part. It's not guest-visible. Even the device model code is (should be!) ignorant of it.

Device guest part reconfiguration at run-time is done by the guest OS, just like for non-virtual devices. This may make the device model call out to the block layer to adjust settings there.

Except for configuration knobs that match physical knobs (media eject and insert, for instance). Those belong into the monitor.

So, if your configuration knob can be manipulated at run time from the monitor, and it doesn't correspond to a physical knob, then it *probably* belongs to the host part, and its setting isn't guest-visible.

Let's assume we're indeed talking about reconfiguring the host part. A generic command for that makes sense to me. Could look like "blockdev_set ID NAME=VAL,..." in the human monitor. The NAME=VAL should be consistent with blockdev_add whenever possible. Requires a crystal ball until we actually get blockdev_add. If we're afraid of getting it wrong, we could do a drive_set now.

#9 lore raw

Re: [Qemu-devel] [V2 2/2]Qemu: Add commands "hostcache_set" and "hostcache_get"

Am 23.05.2011 18:01, schrieb Markus Armbruster:

Quoted 75 lines
Quoted 54 lines · Kevin Wolf

Am 23.05.2011 12:00, schrieb Stefan Hajnoczi:

Quoted 35 lines · Supriya Kannery
Quoted 34 lines · Stefan Hajnoczi
Quoted 32 lines · Supriya Kannery
Quoted 3 lines

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

A generic command for changing block device options would be nice, althought I don't see other options where it makes sense to change them at runtime.

The alternative would be:

block_set hostcache on

"block_set", {"device": "ide1-cd0", "name": "hostcache", "enable": true}

The hostcache_get information would be part of query-block output:

         {
            "device":"ide0-hd0",
            "locked":false,
            "removable":false,
            "inserted":{
               "ro":false,
               "drv":"qcow2",
               "encrypted":false,
               "file":"disks/test.img"
              "hostcache":true,
            },
            "type":"hd"
         },

This approach is extensible if more options need to be exposed.

Quoted 8 lines
Quoted 4 lines

Sure, I will resubmit this patchset, after making this feature more generic. Can you pls help finding atleast one or two options (other than hostcache) which can be changed dynamically. This will help me evaluate the generic approach.

Hang on, let's see if we can get agreement from Kevin and others before taking this approach. Like I said, I don't see other options that should be changed at runtime.

Things like enabling copy on read could fit here.

Generally I'm in favour of having a generic command. We just need to pay attention not to include things that we don't want to maintain long term, i.e. just putting the current cache=... parameter into the argument isn't going to work. Maybe two booleans 'o_direct' and 'ignore_flushes' is what we want to have. The same structure should be used for blkdev_add then, even though it will allow some more options.

Quoted 5 lines

I'm also not completely sure how you would enable cache=writethrough from the command line in a fully converted world. Would this be one of the arguments that must be specified on BlockDriverState creation and can't be changed later, and the device will pick it up from there? Or is it a qdev property and somehow makes it way to the block layer?

qdev properties are the configurable bits of the device's guest part. For something to be a qdev property, it must belong to the guest part, and it must be configurable.

Example: a drive serial number belongs to the guest part. It's guest-visible. The host part doesn't care about it; in fact, if you unplug the device, you could reuse the same host part with another guest part with different serial number.

Example: the image format doesn't belong to the guest part. It's not guest-visible. Even the device model code is (should be!) ignorant of it.

Quoted 6 lines

Device guest part reconfiguration at run-time is done by the guest OS, just like for non-virtual devices. This may make the device model call out to the block layer to adjust settings there.

Except for configuration knobs that match physical knobs (media eject and insert, for instance). Those belong into the monitor.

Thanks Markus. I think I'm well aware what a guest and what a host is. The trouble begins when there's a connection between both...

Let me try to list the facts with cache=writethrough:

* It is currently considered host state (part of bs->open_flags). It results in the image file being opened with O_SYNC.

* It is guest visible as something like a WCE bit (Write Cache Enable). The guest is allowed to toggle this bit on real hardware and we want to add this functionality, too.

* We need a way to configure it from the start. Currently done by cache=writethrough, but that is host state even though it's guest visible. That would be an argument for moving it into qdev.

* You're supposed to use blockdev_add first and then device_add. This means that we open the image file before the device and its qdev config exists. So the first thing the device must do is to reopen the image? What to do with the fd: protocol?

So, treating it as host state is wrong. Treating it as guest state doesn't work really well either.

Quoted 11 lines

So, if your configuration knob can be manipulated at run time from the monitor, and it doesn't correspond to a physical knob, then it *probably* belongs to the host part, and its setting isn't guest-visible.

Let's assume we're indeed talking about reconfiguring the host part. A generic command for that makes sense to me. Could look like "blockdev_set ID NAME=VAL,..." in the human monitor. The NAME=VAL should be consistent with blockdev_add whenever possible. Requires a crystal ball until we actually get blockdev_add. If we're afraid of getting it wrong, we could do a drive_set now.

If we're always afraid of getting it wrong and try to evade, we'll never get blockdev_add.

Kevin

#10 lore raw

Re: [Qemu-devel] [V2 2/2]Qemu: Add commands "hostcache_set" and "hostcache_get"

Quoted 99 lines · Kevin Wolf

Am 23.05.2011 18:01, schrieb Markus Armbruster:

Quoted 75 lines
Quoted 54 lines · Kevin Wolf

Am 23.05.2011 12:00, schrieb Stefan Hajnoczi:

Quoted 35 lines · Supriya Kannery
Quoted 34 lines · Stefan Hajnoczi
Quoted 32 lines · Supriya Kannery
Quoted 3 lines

Monitor commands "hostcache_set" and "hostcache_get" added for dynamic host cache change and display of host cache setting respectively.

A generic command for changing block device options would be nice, althought I don't see other options where it makes sense to change them at runtime.

The alternative would be:

block_set hostcache on

"block_set", {"device": "ide1-cd0", "name": "hostcache", "enable": true}

The hostcache_get information would be part of query-block output:

         {
            "device":"ide0-hd0",
            "locked":false,
            "removable":false,
            "inserted":{
               "ro":false,
               "drv":"qcow2",
               "encrypted":false,
               "file":"disks/test.img"
              "hostcache":true,
            },
            "type":"hd"
         },

This approach is extensible if more options need to be exposed.

Quoted 8 lines
Quoted 4 lines

Sure, I will resubmit this patchset, after making this feature more generic. Can you pls help finding atleast one or two options (other than hostcache) which can be changed dynamically. This will help me evaluate the generic approach.

Hang on, let's see if we can get agreement from Kevin and others before taking this approach. Like I said, I don't see other options that should be changed at runtime.

Things like enabling copy on read could fit here.

Generally I'm in favour of having a generic command. We just need to pay attention not to include things that we don't want to maintain long term, i.e. just putting the current cache=... parameter into the argument isn't going to work. Maybe two booleans 'o_direct' and 'ignore_flushes' is what we want to have. The same structure should be used for blkdev_add then, even though it will allow some more options.

Quoted 5 lines

I'm also not completely sure how you would enable cache=writethrough from the command line in a fully converted world. Would this be one of the arguments that must be specified on BlockDriverState creation and can't be changed later, and the device will pick it up from there? Or is it a qdev property and somehow makes it way to the block layer?

qdev properties are the configurable bits of the device's guest part. For something to be a qdev property, it must belong to the guest part, and it must be configurable.

Example: a drive serial number belongs to the guest part. It's guest-visible. The host part doesn't care about it; in fact, if you unplug the device, you could reuse the same host part with another guest part with different serial number.

Example: the image format doesn't belong to the guest part. It's not guest-visible. Even the device model code is (should be!) ignorant of it.

Quoted 6 lines

Device guest part reconfiguration at run-time is done by the guest OS, just like for non-virtual devices. This may make the device model call out to the block layer to adjust settings there.

Except for configuration knobs that match physical knobs (media eject and insert, for instance). Those belong into the monitor.

Thanks Markus. I think I'm well aware what a guest and what a host is. The trouble begins when there's a connection between both...

Let me try to list the facts with cache=writethrough:

* It is currently considered host state (part of bs->open_flags). It results in the image file being opened with O_SYNC.

* It is guest visible as something like a WCE bit (Write Cache Enable). The guest is allowed to toggle this bit on real hardware and we want to add this functionality, too.

* We need a way to configure it from the start. Currently done by cache=writethrough, but that is host state even though it's guest visible. That would be an argument for moving it into qdev.

Quoted 7 lines

* You're supposed to use blockdev_add first and then device_add. This means that we open the image file before the device and its qdev config exists. So the first thing the device must do is to reopen the image? What to do with the fd: protocol?

So, treating it as host state is wrong. Treating it as guest state doesn't work really well either.

Similar to read-only, except for the "guest can change it" bit.

For read-only, we concluded that it belongs to both guest and host part. Connecting a read/write guest part to a read-only host part is not allowed. For convenience, the guest part's read-only bit could default to match the host part's.

Can we solve the WCE bit problem similarly?

Quoted 12 lines
Quoted 11 lines

So, if your configuration knob can be manipulated at run time from the monitor, and it doesn't correspond to a physical knob, then it *probably* belongs to the host part, and its setting isn't guest-visible.

Let's assume we're indeed talking about reconfiguring the host part. A generic command for that makes sense to me. Could look like "blockdev_set ID NAME=VAL,..." in the human monitor. The NAME=VAL should be consistent with blockdev_add whenever possible. Requires a crystal ball until we actually get blockdev_add. If we're afraid of getting it wrong, we could do a drive_set now.

Quoted 2 lines

If we're always afraid of getting it wrong and try to evade, we'll never get blockdev_add.

Yes. But it's advisable to start with blockdev_add rather than with blockdev_set.