Skip to contentSkip to conversation

Compat props bug fixes

PATCH In review Markus Armbruster 4 participants 20 messages Apr 1, 2019 – Apr 2, 2019

lore
Share & tools
lore
#0 lore raw

v2 cover Compat props bug fixes

v2:
* PATCH 1-4: Complete rewrite of fatally flawed PATCH v1 1/2 [Igor]

Markus Armbruster (5):
Revert "vl: Fix to create migration object before block backends
again"
Revert "migration: move only_migratable to MigrationState"
migration: Support adding migration blockers earlier
vl: Document dependencies hiding in global and compat props
accel: Unbreak accelerator fallback

 accel/accel.c           |  3 ++-
 include/sysemu/sysemu.h |  1 +
 migration/migration.c   | 11 +++++++----
 migration/migration.h   |  3 ---
 migration/savevm.c      |  2 +-
 vl.c                    | 44 +++++++++++++++++++++++++++--------------
 6 files changed, 40 insertions(+), 24 deletions(-)
Signature · 1 line
2.17.2
#1 lore raw

PATCH v2 1/5 Revert "vl: Fix to create migration object before block backends again"

This reverts commit e60483f2f8498ae08ae79ca4c6fb03a3317f5e1e.

Recent commit cda4aa9a5a0 moved block backend creation before machine property evaluation. This broke block backends registering migration blockers. Commit e60483f2f84 fixed it by moving migration object creation before block backend creation. This broke migration with Xen. Turns out we need to configure the accelerator before we create the migration object so that Xen's accelerator compat properties get applied. Revert the flawed commit. This fixes the Xen regression, but brings back the block backend regression. The next commits will fix it again.

---
 vl.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)
vl.c +7 -8
index d61d5604e5..c1d5484e12 100644
@@ -4276,17 +4276,10 @@ int main(int argc, char **argv, char **envp)
         exit(0);
     }
 
-    /*
-     * Migration object can only be created after global properties
-     * are applied correctly.
-     */
-    migration_object_init();
-
     /*
      * Note: we need to create block backends before
      * machine_set_property(), so machine properties can refer to
-     * them, and after migration_object_init(), so we can create
-     * migration blockers.
+     * them.
      */
     configure_blockdev(&bdo_queue, machine_class, snapshot);
 
@@ -4304,6 +4297,12 @@ int main(int argc, char **argv, char **envp)
                      machine_class->name, machine_class->deprecation_reason);
     }
 
+    /*
+     * Migration object can only be created after global properties
+     * are applied correctly.
+     */
+    migration_object_init();
+
     if (qtest_chrdev) {
         qtest_init(qtest_chrdev, qtest_log, &error_fatal);
     }
Signature · 1 line
2.17.2
Reviewed-by on v2 1/5
#2 lore raw

Re: [Qemu-devel] [PATCH v2 1/5] Revert "vl: Fix to create migration object before block backends again"

Quoted 4 lines · On Mon, 1 Apr 2019 11:08:23 +0200 Markus Armbruster <[email protected]> wrote:

This reverts commit e60483f2f8498ae08ae79ca4c6fb03a3317f5e1e.

Recent commit cda4aa9a5a0 moved block backend creation before machine property evaluation.

Side note:
I've also looked at
"[PATCH v3 11/12] pc: Support firmware configuration with -blockdev"
that caused movement and have a similar issue with hostmem backends where I wanted to set machine.inital-ram-memdev property pointing to backend. But hostmem backends are created, since they have to be created after machine_set_properties/configure_accelerator due to dependencies.

After some playing around with link property on machine and trying to untangle dependency chain I gave up (doesn't look feasible). Idea with aliasing like with blockdev won't work either due to dependencies (also it looks to me to be quite complicated one). So I've resorted to "usual" machine practice 'machine.set_string_property' and deal with 'id' referred by it later at machine_run_board_init() time to instantiate initial RAM, when we already created all backends.

Quoted 10 lines

This broke block backends registering migration
blockers. Commit e60483f2f84 fixed it by moving migration object creation before block backend creation. This broke migration with Xen. Turns out we need to configure the accelerator before we create the migration object so that Xen's accelerator compat properties get applied. Revert the flawed commit. This fixes the Xen regression, but brings back the block backend regression. The next commits will fix it again.

Quoted 40 lines
---
 vl.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)
vl.c +7 -8
index d61d5604e5..c1d5484e12 100644
@@ -4276,17 +4276,10 @@ int main(int argc, char **argv, char **envp)
         exit(0);
     }
 
-    /*
-     * Migration object can only be created after global properties
-     * are applied correctly.
-     */
-    migration_object_init();
-
     /*
      * Note: we need to create block backends before
      * machine_set_property(), so machine properties can refer to
-     * them, and after migration_object_init(), so we can create
-     * migration blockers.
+     * them.
      */
     configure_blockdev(&bdo_queue, machine_class, snapshot);
 
@@ -4304,6 +4297,12 @@ int main(int argc, char **argv, char **envp)
                      machine_class->name, machine_class->deprecation_reason);
     }
 
+    /*
+     * Migration object can only be created after global properties
+     * are applied correctly.
+     */
+    migration_object_init();
+
     if (qtest_chrdev) {
         qtest_init(qtest_chrdev, qtest_log, &error_fatal);
     }
#3 lore raw

Re: [Qemu-devel] [PATCH v2 1/5] Revert "vl: Fix to create migration object before block backends again"

Quoted 16 lines · Igor Mammedov
Quoted 4 lines · On Mon, 1 Apr 2019 11:08:23 +0200 Markus Armbruster <[email protected]> wrote:

This reverts commit e60483f2f8498ae08ae79ca4c6fb03a3317f5e1e.

Recent commit cda4aa9a5a0 moved block backend creation before machine property evaluation.

Side note:
I've also looked at
"[PATCH v3 11/12] pc: Support firmware configuration with -blockdev"
that caused movement and have a similar issue with hostmem backends where I wanted to set machine.inital-ram-memdev property pointing to backend. But hostmem backends are created, since they have to be created after machine_set_properties/configure_accelerator due to dependencies.

Quoted 7 lines

After some playing around with link property on machine and trying to untangle dependency chain I gave up (doesn't look feasible). Idea with aliasing like with blockdev won't work either due to dependencies (also it looks to me to be quite complicated one). So I've resorted to "usual" machine practice 'machine.set_string_property' and deal with 'id' referred by it later at machine_run_board_init() time to instantiate initial RAM, when we already created all backends.

That's... horrible. How many of these monstrosities have crept in?

Let's discuss once the RC circus has moved to the next town.

Quoted 9 lines
Quoted 9 lines

This broke block backends registering migration
blockers. Commit e60483f2f84 fixed it by moving migration object creation before block backend creation. This broke migration with Xen. Turns out we need to configure the accelerator before we create the migration object so that Xen's accelerator compat properties get applied. Revert the flawed commit. This fixes the Xen regression, but brings back the block backend regression. The next commits will fix it again.

Quoted 3 lines
Quoted 1 line

Thanks!

#4 lore raw

PATCH v2 2/5 Revert "migration: move only_migratable to MigrationState"

In reply to Markus Armbruster

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

---
 include/sysemu/sysemu.h | 1 +
 migration/migration.c   | 5 ++---
 migration/migration.h   | 3 ---
 migration/savevm.c      | 2 +-
 vl.c                    | 9 ++-------
 5 files changed, 6 insertions(+), 14 deletions(-)
include/sysemu/sysemu.h +1 -0
index 6065d9e420..5f133cae83 100644
@@ -14,6 +14,7 @@
 /* vl.c */
 
 extern const char *bios_name;
+extern int only_migratable;
 extern const char *qemu_name;
 extern QemuUUID qemu_uuid;
 extern bool qemu_uuid_set;
migration/migration.c +2 -3
index 69f75124c9..f6076e5295 100644
@@ -1707,7 +1707,7 @@ static GSList *migration_blockers;
 
 int migrate_add_blocker(Error *reason, Error **errp)
 {
-    if (migrate_get_current()->only_migratable) {
+    if (only_migratable) {
         error_propagate_prepend(errp, error_copy(reason),
                                 "disallowing migration blocker "
                                 "(--only_migratable) for: ");
@@ -3337,7 +3337,7 @@ void migration_global_dump(Monitor *mon)
     monitor_printf(mon, "store-global-state: %s\n",
                    ms->store_global_state ? "on" : "off");
     monitor_printf(mon, "only-migratable: %s\n",
-                   ms->only_migratable ? "on" : "off");
+                   only_migratable ? "on" : "off");
     monitor_printf(mon, "send-configuration: %s\n",
                    ms->send_configuration ? "on" : "off");
     monitor_printf(mon, "send-section-footer: %s\n",
@@ -3352,7 +3352,6 @@ void migration_global_dump(Monitor *mon)
 static Property migration_properties[] = {
     DEFINE_PROP_BOOL("store-global-state", MigrationState,
                      store_global_state, true),
-    DEFINE_PROP_BOOL("only-migratable", MigrationState, only_migratable, false),
     DEFINE_PROP_BOOL("send-configuration", MigrationState,
                      send_configuration, true),
     DEFINE_PROP_BOOL("send-section-footer", MigrationState,
migration/migration.h +0 -3
index 0f986935e1..438f17edad 100644
@@ -219,9 +219,6 @@ struct MigrationState
      */
     bool store_global_state;
 
-    /* Whether the VM is only allowing for migratable devices */
-    bool only_migratable;
-
     /* Whether we send QEMU_VM_CONFIGURATION during migration */
     bool send_configuration;
     /* Whether we send section footer during migration */
migration/savevm.c +1 -1
index 1415001d1c..34bcad3807 100644
@@ -2844,7 +2844,7 @@ void vmstate_register_ram_global(MemoryRegion *mr)
 bool vmstate_check_only_migratable(const VMStateDescription *vmsd)
 {
     /* check needed if --only-migratable is specified */
-    if (!migrate_get_current()->only_migratable) {
+    if (!only_migratable) {
         return true;
     }
 
vl.c +2 -7
index c1d5484e12..e4d7ad6b85 100644
@@ -185,6 +185,7 @@ const char *prom_envs[MAX_PROM_ENVS];
 int boot_menu;
 bool boot_strict;
 uint8_t *boot_splash_filedata;
+int only_migratable; /* turn it off unless user states otherwise */
 bool wakeup_suspend_enabled;
 
 int icount_align_option;
@@ -3799,13 +3800,7 @@ int main(int argc, char **argv, char **envp)
                 incoming = optarg;
                 break;
             case QEMU_OPTION_only_migratable:
-                /*
-                 * TODO: we can remove this option one day, and we
-                 * should all use:
-                 *
-                 * "-global migration.only-migratable=true"
-                 */
-                qemu_global_option("migration.only-migratable=true");
+                only_migratable = 1;
                 break;
             case QEMU_OPTION_nodefaults:
                 has_defaults = 0;
Signature · 1 line
2.17.2
#5 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] Revert "migration: move only_migratable to MigrationState"

Quoted 110 lines · * Markus Armbruster ([email protected]) wrote:

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

---
 include/sysemu/sysemu.h | 1 +
 migration/migration.c   | 5 ++---
 migration/migration.h   | 3 ---
 migration/savevm.c      | 2 +-
 vl.c                    | 9 ++-------
 5 files changed, 6 insertions(+), 14 deletions(-)
include/sysemu/sysemu.h +1 -0
index 6065d9e420..5f133cae83 100644
@@ -14,6 +14,7 @@
 /* vl.c */
 
 extern const char *bios_name;
+extern int only_migratable;
 extern const char *qemu_name;
 extern QemuUUID qemu_uuid;
 extern bool qemu_uuid_set;
migration/migration.c +2 -2
index 69f75124c9..f6076e5295 100644
@@ -1707,7 +1707,7 @@ static GSList *migration_blockers;
 
 int migrate_add_blocker(Error *reason, Error **errp)
 {
-    if (migrate_get_current()->only_migratable) {
+    if (only_migratable) {
         error_propagate_prepend(errp, error_copy(reason),
                                 "disallowing migration blocker "
                                 "(--only_migratable) for: ");
@@ -3337,7 +3337,7 @@ void migration_global_dump(Monitor *mon)
     monitor_printf(mon, "store-global-state: %s\n",
                    ms->store_global_state ? "on" : "off");
     monitor_printf(mon, "only-migratable: %s\n",
-                   ms->only_migratable ? "on" : "off");
+                   only_migratable ? "on" : "off");
     monitor_printf(mon, "send-configuration: %s\n",
                    ms->send_configuration ? "on" : "off");
     monitor_printf(mon, "send-section-footer: %s\n",

migration/migration.c void migration_global_dump(Monitor *mon) · line 3355

Quoted 4 lines
migration/migration.c +0 -1
@@ -3352,7 +3352,6 @@ void migration_global_dump(Monitor *mon)
 static Property migration_properties[] = {
     DEFINE_PROP_BOOL("store-global-state", MigrationState,
                      store_global_state, true),
-    DEFINE_PROP_BOOL("only-migratable", MigrationState, only_migratable, false),

So I'm generally OK at this patch, but I'm worried whether this is API?

Dave

Quoted 60 lines
migration/migration.c +0 -0
     DEFINE_PROP_BOOL("send-configuration", MigrationState,
                      send_configuration, true),
     DEFINE_PROP_BOOL("send-section-footer", MigrationState,
migration/migration.h +0 -3
index 0f986935e1..438f17edad 100644
@@ -219,9 +219,6 @@ struct MigrationState
      */
     bool store_global_state;
 
-    /* Whether the VM is only allowing for migratable devices */
-    bool only_migratable;
-
     /* Whether we send QEMU_VM_CONFIGURATION during migration */
     bool send_configuration;
     /* Whether we send section footer during migration */
migration/savevm.c +1 -1
index 1415001d1c..34bcad3807 100644
@@ -2844,7 +2844,7 @@ void vmstate_register_ram_global(MemoryRegion *mr)
 bool vmstate_check_only_migratable(const VMStateDescription *vmsd)
 {
     /* check needed if --only-migratable is specified */
-    if (!migrate_get_current()->only_migratable) {
+    if (!only_migratable) {
         return true;
     }
 
vl.c +2 -7
index c1d5484e12..e4d7ad6b85 100644
@@ -185,6 +185,7 @@ const char *prom_envs[MAX_PROM_ENVS];
 int boot_menu;
 bool boot_strict;
 uint8_t *boot_splash_filedata;
+int only_migratable; /* turn it off unless user states otherwise */
 bool wakeup_suspend_enabled;
 
 int icount_align_option;
@@ -3799,13 +3800,7 @@ int main(int argc, char **argv, char **envp)
                 incoming = optarg;
                 break;
             case QEMU_OPTION_only_migratable:
-                /*
-                 * TODO: we can remove this option one day, and we
-                 * should all use:
-                 *
-                 * "-global migration.only-migratable=true"
-                 */
-                qemu_global_option("migration.only-migratable=true");
+                only_migratable = 1;
                 break;
             case QEMU_OPTION_nodefaults:
                 has_defaults = 0;
Signature · 1 line
2.17.2

--
Dr. David Alan Gilbert / [email protected] / Manchester, UK

#6 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] Revert "migration: move only_migratable to MigrationState"

Quoted 110 lines · Dr. David Alan Gilbert
Quoted 109 lines · * Markus Armbruster ([email protected]) wrote:

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

---
 include/sysemu/sysemu.h | 1 +
 migration/migration.c   | 5 ++---
 migration/migration.h   | 3 ---
 migration/savevm.c      | 2 +-
 vl.c                    | 9 ++-------
 5 files changed, 6 insertions(+), 14 deletions(-)
include/sysemu/sysemu.h +1 -0
index 6065d9e420..5f133cae83 100644
@@ -14,6 +14,7 @@
 /* vl.c */
 
 extern const char *bios_name;
+extern int only_migratable;
 extern const char *qemu_name;
 extern QemuUUID qemu_uuid;
 extern bool qemu_uuid_set;
migration/migration.c +2 -2
index 69f75124c9..f6076e5295 100644
@@ -1707,7 +1707,7 @@ static GSList *migration_blockers;
 
 int migrate_add_blocker(Error *reason, Error **errp)
 {
-    if (migrate_get_current()->only_migratable) {
+    if (only_migratable) {
         error_propagate_prepend(errp, error_copy(reason),
                                 "disallowing migration blocker "
                                 "(--only_migratable) for: ");
@@ -3337,7 +3337,7 @@ void migration_global_dump(Monitor *mon)
     monitor_printf(mon, "store-global-state: %s\n",
                    ms->store_global_state ? "on" : "off");
     monitor_printf(mon, "only-migratable: %s\n",
-                   ms->only_migratable ? "on" : "off");
+                   only_migratable ? "on" : "off");
     monitor_printf(mon, "send-configuration: %s\n",
                    ms->send_configuration ? "on" : "off");
     monitor_printf(mon, "send-section-footer: %s\n",
Quoted 8 lines

migration/migration.c void migration_global_dump(Monitor *mon) · line 3355

Quoted 5 lines
migration/migration.c +0 -1
@@ -3352,7 +3352,6 @@ void migration_global_dump(Monitor *mon)
 static Property migration_properties[] = {
     DEFINE_PROP_BOOL("store-global-state", MigrationState,
                      store_global_state, true),
-    DEFINE_PROP_BOOL("only-migratable", MigrationState, only_migratable, false),

So I'm generally OK at this patch, but I'm worried whether this is API?

IIUC, old code can use either

-global migration.only-migratable=true
-only-migratable

After this patch only

-only-migratable

is valid. So from that POV it is an API break.

Quoted 1 line

From libvirt's POV this is a non-issue as we never use either of these options.

We did have a BZ filed to support it, but never did.

To avoid the API break I guess there would been to be a special case early loop iterating over argv[] looking for '-global migration.only-migratable=true', making that set the static 'bool only_migratable', instead of letting it get handled by the object infra.

Quoted 54 lines
Quoted 54 lines
migration/migration.c +0 -0
     DEFINE_PROP_BOOL("send-configuration", MigrationState,
                      send_configuration, true),
     DEFINE_PROP_BOOL("send-section-footer", MigrationState,
migration/migration.h +0 -3
index 0f986935e1..438f17edad 100644
@@ -219,9 +219,6 @@ struct MigrationState
      */
     bool store_global_state;
 
-    /* Whether the VM is only allowing for migratable devices */
-    bool only_migratable;
-
     /* Whether we send QEMU_VM_CONFIGURATION during migration */
     bool send_configuration;
     /* Whether we send section footer during migration */
migration/savevm.c +1 -1
index 1415001d1c..34bcad3807 100644
@@ -2844,7 +2844,7 @@ void vmstate_register_ram_global(MemoryRegion *mr)
 bool vmstate_check_only_migratable(const VMStateDescription *vmsd)
 {
     /* check needed if --only-migratable is specified */
-    if (!migrate_get_current()->only_migratable) {
+    if (!only_migratable) {
         return true;
     }
 
vl.c +2 -7
index c1d5484e12..e4d7ad6b85 100644
@@ -185,6 +185,7 @@ const char *prom_envs[MAX_PROM_ENVS];
 int boot_menu;
 bool boot_strict;
 uint8_t *boot_splash_filedata;
+int only_migratable; /* turn it off unless user states otherwise */
 bool wakeup_suspend_enabled;
 
 int icount_align_option;
@@ -3799,13 +3800,7 @@ int main(int argc, char **argv, char **envp)
                 incoming = optarg;
                 break;
             case QEMU_OPTION_only_migratable:
-                /*
-                 * TODO: we can remove this option one day, and we
-                 * should all use:
-                 *
-                 * "-global migration.only-migratable=true"
-                 */
-                qemu_global_option("migration.only-migratable=true");
+                only_migratable = 1;
Quoted 9 lines
Quoted 6 lines
vl.c +0 -0
                 break;
             case QEMU_OPTION_nodefaults:
                 has_defaults = 0;
Signature · 1 line
2.17.2

--
Dr. David Alan Gilbert / [email protected] / Manchester, UK

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#7 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] Revert "migration: move only_migratable to MigrationState"

Quoted 111 lines · Daniel P. Berrangé
Quoted 110 lines · Dr. David Alan Gilbert
Quoted 109 lines · * Markus Armbruster ([email protected]) wrote:

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

---
 include/sysemu/sysemu.h | 1 +
 migration/migration.c   | 5 ++---
 migration/migration.h   | 3 ---
 migration/savevm.c      | 2 +-
 vl.c                    | 9 ++-------
 5 files changed, 6 insertions(+), 14 deletions(-)
include/sysemu/sysemu.h +1 -0
index 6065d9e420..5f133cae83 100644
@@ -14,6 +14,7 @@
 /* vl.c */
 
 extern const char *bios_name;
+extern int only_migratable;
 extern const char *qemu_name;
 extern QemuUUID qemu_uuid;
 extern bool qemu_uuid_set;
migration/migration.c +2 -2
index 69f75124c9..f6076e5295 100644
@@ -1707,7 +1707,7 @@ static GSList *migration_blockers;
 
 int migrate_add_blocker(Error *reason, Error **errp)
 {
-    if (migrate_get_current()->only_migratable) {
+    if (only_migratable) {
         error_propagate_prepend(errp, error_copy(reason),
                                 "disallowing migration blocker "
                                 "(--only_migratable) for: ");
@@ -3337,7 +3337,7 @@ void migration_global_dump(Monitor *mon)
     monitor_printf(mon, "store-global-state: %s\n",
                    ms->store_global_state ? "on" : "off");
     monitor_printf(mon, "only-migratable: %s\n",
-                   ms->only_migratable ? "on" : "off");
+                   only_migratable ? "on" : "off");
     monitor_printf(mon, "send-configuration: %s\n",
                    ms->send_configuration ? "on" : "off");
     monitor_printf(mon, "send-section-footer: %s\n",
Quoted 8 lines
Quoted 8 lines

migration/migration.c void migration_global_dump(Monitor *mon) · line 3355

Quoted 5 lines
migration/migration.c +0 -1
@@ -3352,7 +3352,6 @@ void migration_global_dump(Monitor *mon)
 static Property migration_properties[] = {
     DEFINE_PROP_BOOL("store-global-state", MigrationState,
                      store_global_state, true),
-    DEFINE_PROP_BOOL("only-migratable", MigrationState, only_migratable, false),

So I'm generally OK at this patch, but I'm worried whether this is API?

Uh, forgot to discuss this in my commit message, sorry!

Quoted 11 lines

IIUC, old code can use either

-global migration.only-migratable=true
-only-migratable

After this patch only

-only-migratable

is valid. So from that POV it is an API break.

Quoted 7 lines

From libvirt's POV this is a non-issue as we never use either of these options. We did have a BZ filed to support it, but never did.

To avoid the API break I guess there would been to be a special case early loop iterating over argv[] looking for '-global migration.only-migratable=true', making that set the static 'bool only_migratable', instead of letting it get handled by the object infra.

Is this is worth the trouble?

-global migration.only-migratable=on is entirely undocumented. The property appears in -device migration,help and device-list-properties, but that's all.

Why would any sane person risk using arcane & undocumented -global migration.only-migratable=on instead of the obvious & documented -only-migratable?

[...]

#8 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] Revert "migration: move only_migratable to MigrationState"

Quoted 138 lines · Markus Armbruster
Quoted 111 lines · Daniel P. Berrangé
Quoted 110 lines · Dr. David Alan Gilbert
Quoted 109 lines · * Markus Armbruster ([email protected]) wrote:

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

---
 include/sysemu/sysemu.h | 1 +
 migration/migration.c   | 5 ++---
 migration/migration.h   | 3 ---
 migration/savevm.c      | 2 +-
 vl.c                    | 9 ++-------
 5 files changed, 6 insertions(+), 14 deletions(-)
include/sysemu/sysemu.h +1 -0
index 6065d9e420..5f133cae83 100644
@@ -14,6 +14,7 @@
 /* vl.c */
 
 extern const char *bios_name;
+extern int only_migratable;
 extern const char *qemu_name;
 extern QemuUUID qemu_uuid;
 extern bool qemu_uuid_set;
migration/migration.c +2 -2
index 69f75124c9..f6076e5295 100644
@@ -1707,7 +1707,7 @@ static GSList *migration_blockers;
 
 int migrate_add_blocker(Error *reason, Error **errp)
 {
-    if (migrate_get_current()->only_migratable) {
+    if (only_migratable) {
         error_propagate_prepend(errp, error_copy(reason),
                                 "disallowing migration blocker "
                                 "(--only_migratable) for: ");
@@ -3337,7 +3337,7 @@ void migration_global_dump(Monitor *mon)
     monitor_printf(mon, "store-global-state: %s\n",
                    ms->store_global_state ? "on" : "off");
     monitor_printf(mon, "only-migratable: %s\n",
-                   ms->only_migratable ? "on" : "off");
+                   only_migratable ? "on" : "off");
     monitor_printf(mon, "send-configuration: %s\n",
                    ms->send_configuration ? "on" : "off");
     monitor_printf(mon, "send-section-footer: %s\n",
Quoted 8 lines
Quoted 8 lines

migration/migration.c void migration_global_dump(Monitor *mon) · line 3355

Quoted 5 lines
migration/migration.c +0 -1
@@ -3352,7 +3352,6 @@ void migration_global_dump(Monitor *mon)
 static Property migration_properties[] = {
     DEFINE_PROP_BOOL("store-global-state", MigrationState,
                      store_global_state, true),
-    DEFINE_PROP_BOOL("only-migratable", MigrationState, only_migratable, false),

So I'm generally OK at this patch, but I'm worried whether this is API?

Uh, forgot to discuss this in my commit message, sorry!

Quoted 14 lines

IIUC, old code can use either

-global migration.only-migratable=true
-only-migratable

After this patch only

-only-migratable

is valid. So from that POV it is an API break.

From libvirt's POV this is a non-issue as we never use either of these options. We did have a BZ filed to support it, but never did.

Quoted 6 lines
Quoted 4 lines

To avoid the API break I guess there would been to be a special case early loop iterating over argv[] looking for '-global migration.only-migratable=true', making that set the static 'bool only_migratable', instead of letting it get handled by the object infra.

Is this is worth the trouble?

Probably not.

Quoted 7 lines

-global migration.only-migratable=on is entirely undocumented. The property appears in -device migration,help and device-list-properties, but that's all.

Why would any sane person risk using arcane & undocumented -global migration.only-migratable=on instead of the obvious & documented -only-migratable?

Given the combination of no known users, current broken impl, and its existance not documented, I think there's a reasonable case for just removing it.

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#9 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] Revert "migration: move only_migratable to MigrationState"

Quoted 152 lines · * Daniel P. Berrangé ([email protected]) wrote:
Quoted 138 lines · Markus Armbruster
Quoted 111 lines · Daniel P. Berrangé
Quoted 110 lines · Dr. David Alan Gilbert
Quoted 109 lines · * Markus Armbruster ([email protected]) wrote:

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

---
 include/sysemu/sysemu.h | 1 +
 migration/migration.c   | 5 ++---
 migration/migration.h   | 3 ---
 migration/savevm.c      | 2 +-
 vl.c                    | 9 ++-------
 5 files changed, 6 insertions(+), 14 deletions(-)
include/sysemu/sysemu.h +1 -0
index 6065d9e420..5f133cae83 100644
@@ -14,6 +14,7 @@
 /* vl.c */
 
 extern const char *bios_name;
+extern int only_migratable;
 extern const char *qemu_name;
 extern QemuUUID qemu_uuid;
 extern bool qemu_uuid_set;
migration/migration.c +2 -2
index 69f75124c9..f6076e5295 100644
@@ -1707,7 +1707,7 @@ static GSList *migration_blockers;
 
 int migrate_add_blocker(Error *reason, Error **errp)
 {
-    if (migrate_get_current()->only_migratable) {
+    if (only_migratable) {
         error_propagate_prepend(errp, error_copy(reason),
                                 "disallowing migration blocker "
                                 "(--only_migratable) for: ");
@@ -3337,7 +3337,7 @@ void migration_global_dump(Monitor *mon)
     monitor_printf(mon, "store-global-state: %s\n",
                    ms->store_global_state ? "on" : "off");
     monitor_printf(mon, "only-migratable: %s\n",
-                   ms->only_migratable ? "on" : "off");
+                   only_migratable ? "on" : "off");
     monitor_printf(mon, "send-configuration: %s\n",
                    ms->send_configuration ? "on" : "off");
     monitor_printf(mon, "send-section-footer: %s\n",
Quoted 8 lines
Quoted 8 lines

migration/migration.c void migration_global_dump(Monitor *mon) · line 3355

Quoted 5 lines
migration/migration.c +0 -1
@@ -3352,7 +3352,6 @@ void migration_global_dump(Monitor *mon)
 static Property migration_properties[] = {
     DEFINE_PROP_BOOL("store-global-state", MigrationState,
                      store_global_state, true),
-    DEFINE_PROP_BOOL("only-migratable", MigrationState, only_migratable, false),

So I'm generally OK at this patch, but I'm worried whether this is API?

Uh, forgot to discuss this in my commit message, sorry!

Quoted 14 lines

IIUC, old code can use either

-global migration.only-migratable=true
-only-migratable

After this patch only

-only-migratable

is valid. So from that POV it is an API break.

From libvirt's POV this is a non-issue as we never use either of these options. We did have a BZ filed to support it, but never did.

Quoted 6 lines
Quoted 4 lines

To avoid the API break I guess there would been to be a special case early loop iterating over argv[] looking for '-global migration.only-migratable=true', making that set the static 'bool only_migratable', instead of letting it get handled by the object infra.

Is this is worth the trouble?

Probably not.

Quoted 4 lines

-global migration.only-migratable=on is entirely undocumented. The property appears in -device migration,help and device-list-properties, but that's all.

Quoted 7 lines
Quoted 3 lines

Why would any sane person risk using arcane & undocumented -global migration.only-migratable=on instead of the obvious & documented -only-migratable?

Given the combination of no known users, current broken impl, and its existance not documented, I think there's a reasonable case for just removing it.

Yes, agreed - it should be noted though.

Dave

Quoted 6 lines

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|

--
Dr. David Alan Gilbert / [email protected] / Manchester, UK

Reviewed-by on v2 2/5
#10 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] Revert "migration: move only_migratable to MigrationState"

In reply to Markus Armbruster
Quoted 60 lines · On Mon, 1 Apr 2019 11:08:24 +0200 Markus Armbruster <[email protected]> wrote:

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Quoted 7 lines

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

with commit message amended like mentioned by David (wrt removed "only-migratable" property)

Quoted 104 lines
---
 include/sysemu/sysemu.h | 1 +
 migration/migration.c   | 5 ++---
 migration/migration.h   | 3 ---
 migration/savevm.c      | 2 +-
 vl.c                    | 9 ++-------
 5 files changed, 6 insertions(+), 14 deletions(-)
include/sysemu/sysemu.h +1 -0
index 6065d9e420..5f133cae83 100644
@@ -14,6 +14,7 @@
 /* vl.c */
 
 extern const char *bios_name;
+extern int only_migratable;
 extern const char *qemu_name;
 extern QemuUUID qemu_uuid;
 extern bool qemu_uuid_set;
migration/migration.c +2 -3
index 69f75124c9..f6076e5295 100644
@@ -1707,7 +1707,7 @@ static GSList *migration_blockers;
 
 int migrate_add_blocker(Error *reason, Error **errp)
 {
-    if (migrate_get_current()->only_migratable) {
+    if (only_migratable) {
         error_propagate_prepend(errp, error_copy(reason),
                                 "disallowing migration blocker "
                                 "(--only_migratable) for: ");
@@ -3337,7 +3337,7 @@ void migration_global_dump(Monitor *mon)
     monitor_printf(mon, "store-global-state: %s\n",
                    ms->store_global_state ? "on" : "off");
     monitor_printf(mon, "only-migratable: %s\n",
-                   ms->only_migratable ? "on" : "off");
+                   only_migratable ? "on" : "off");
     monitor_printf(mon, "send-configuration: %s\n",
                    ms->send_configuration ? "on" : "off");
     monitor_printf(mon, "send-section-footer: %s\n",
@@ -3352,7 +3352,6 @@ void migration_global_dump(Monitor *mon)
 static Property migration_properties[] = {
     DEFINE_PROP_BOOL("store-global-state", MigrationState,
                      store_global_state, true),
-    DEFINE_PROP_BOOL("only-migratable", MigrationState, only_migratable, false),
     DEFINE_PROP_BOOL("send-configuration", MigrationState,
                      send_configuration, true),
     DEFINE_PROP_BOOL("send-section-footer", MigrationState,
migration/migration.h +0 -3
index 0f986935e1..438f17edad 100644
@@ -219,9 +219,6 @@ struct MigrationState
      */
     bool store_global_state;
 
-    /* Whether the VM is only allowing for migratable devices */
-    bool only_migratable;
-
     /* Whether we send QEMU_VM_CONFIGURATION during migration */
     bool send_configuration;
     /* Whether we send section footer during migration */
migration/savevm.c +1 -1
index 1415001d1c..34bcad3807 100644
@@ -2844,7 +2844,7 @@ void vmstate_register_ram_global(MemoryRegion *mr)
 bool vmstate_check_only_migratable(const VMStateDescription *vmsd)
 {
     /* check needed if --only-migratable is specified */
-    if (!migrate_get_current()->only_migratable) {
+    if (!only_migratable) {
         return true;
     }
 
vl.c +2 -7
index c1d5484e12..e4d7ad6b85 100644
@@ -185,6 +185,7 @@ const char *prom_envs[MAX_PROM_ENVS];
 int boot_menu;
 bool boot_strict;
 uint8_t *boot_splash_filedata;
+int only_migratable; /* turn it off unless user states otherwise */
 bool wakeup_suspend_enabled;
 
 int icount_align_option;
@@ -3799,13 +3800,7 @@ int main(int argc, char **argv, char **envp)
                 incoming = optarg;
                 break;
             case QEMU_OPTION_only_migratable:
-                /*
-                 * TODO: we can remove this option one day, and we
-                 * should all use:
-                 *
-                 * "-global migration.only-migratable=true"
-                 */
-                qemu_global_option("migration.only-migratable=true");
+                only_migratable = 1;
                 break;
             case QEMU_OPTION_nodefaults:
                 has_defaults = 0;
#11 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] Revert "migration: move only_migratable to MigrationState"

Quoted 69 lines · Igor Mammedov
Quoted 66 lines · On Mon, 1 Apr 2019 11:08:24 +0200 Markus Armbruster <[email protected]> wrote:

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

Quoted 4 lines
Quoted 1 line

with commit message amended like mentioned by David (wrt removed "only-migratable" property)

Inserting right before the "Conflicts:" line:

Note that the reverted commit made -only-migratable sugar for -global migration.only-migratable=on below the hood. Documentation has only ever mentioned -only-migratable. This commit removes the arcane & undocumented alternative to -only-migratable again. Nobody should be using it.

Quoted 1 line

Thanks!

#12 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] Revert "migration: move only_migratable to MigrationState"

Quoted 76 lines · * Markus Armbruster ([email protected]) wrote:
Quoted 69 lines · Igor Mammedov
Quoted 66 lines · On Mon, 1 Apr 2019 11:08:24 +0200 Markus Armbruster <[email protected]> wrote:

This reverts commit 3df663e575f1876d7f3bc684f80e72fca0703d39. This reverts commit b605c47b57b58e61a901a50a0762dccf43d94783.

Command line option --only-migratable is for disallowing any configuration that can block migration.

Initially, --only-migratable set global variable @only_migratable.

Commit 3df663e575 "migration: move only_migratable to MigrationState" replaced it by MigrationState member @only_migratable. That was a mistake.

First, it doesn't make sense on the design level. MigrationState captures the state of an individual migration, but --only-migratable isn't a property of an individual migration, it's a restriction on QEMU configuration. With fault tolerance, we could have several migrations at once. --only-migratable would certainly protect all of them. Storing it in MigrationState feels inappropriate.

Second, it contributes to a dependency cycle that manifests itself as a bug now.

Putting @only_migratable into MigrationState means its available only after migration_object_init().

We can't set it before migration_object_init(), so we delay setting it with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").

We can't get it before migration_object_init(), so anything that uses it can only run afterwards.

Since migrate_add_blocker() needs to obey --only-migratable, any code adding migration blockers can run only afterwards. This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers.

Moving @only_migratable into MigrationState was a mistake. Revert it.

This doesn't quite break the "migration_object_init() before configure_blockdev() dependency, since migrate_add_blocker() still has another dependency on migration_object_init(). To be addressed the next commit

Conflicts:
include/migration/misc.h
migration/migration.c
migration/migration.h
vl.c

Quoted 4 lines
Quoted 1 line

with commit message amended like mentioned by David (wrt removed "only-migratable" property)

Quoted 7 lines

Inserting right before the "Conflicts:" line:

Note that the reverted commit made -only-migratable sugar for -global migration.only-migratable=on below the hood. Documentation has only ever mentioned -only-migratable. This commit removes the arcane & undocumented alternative to -only-migratable again. Nobody should be using it.

Yes, that's fine:

Quoted 3 lines
Quoted 1 line

Thanks!

--
Dr. David Alan Gilbert / [email protected] / Manchester, UK

#13 lore raw

PATCH v2 3/5 migration: Support adding migration blockers earlier

In reply to Markus Armbruster

migrate_add_blocker() asserts we have a current_migration object, in migrate_get_current(). We do only after migration_object_init().

This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers, as demonstrated by qemu-iotests 055.

To fix it, break the last dependency: make migrate_add_blocker() usable before migration_object_init().

The previous commit already removed the use of migrate_get_current() from migrate_add_blocker() itself. Didn't quite do the trick, as there's another one hiding in migration_is_idle().

The use there isn't actually necessary: when no migration object has been created yet, migration is surely idle. Make migration_is_idle() return true then.

Fixes cda4aa9a5a08777cf13e164c0543bd4888b8adce
---
 migration/migration.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
migration/migration.c +5 -1
index f6076e5295..609e0df5d0 100644
@@ -1646,7 +1646,11 @@ bool migration_in_postcopy_after_devices(MigrationState *s)
 
 bool migration_is_idle(void)
 {
-    MigrationState *s = migrate_get_current();
+    MigrationState *s = current_migration;
+
+    if (!s) {
+        return true;
+    }
 
     switch (s->state) {
     case MIGRATION_STATUS_NONE:
Signature · 1 line
2.17.2
Reviewed-by on v2 3/5
#14 lore raw

Re: [Qemu-devel] [PATCH v2 3/5] migration: Support adding migration blockers earlier

Quoted 30 lines · On Mon, 1 Apr 2019 11:08:25 +0200 Markus Armbruster <[email protected]> wrote:

migrate_add_blocker() asserts we have a current_migration object, in migrate_get_current(). We do only after migration_object_init().

This contributes to the following dependency cycle:

* configure_blockdev() must run before machine_set_property() so machine properties can refer to block backends

* machine_set_property() before configure_accelerator() so machine properties like kvm-irqchip get applied

* configure_accelerator() before migration_object_init() so that Xen's accelerator compat properties get applied.

* migration_object_init() before configure_blockdev() so configure_blockdev() can add migration blockers

The cycle was closed when recent commit cda4aa9a5a0 "Create block backends before setting machine properties" added the first dependency, and satisfied it by violating the last one. Broke block backends that add migration blockers, as demonstrated by qemu-iotests 055.

To fix it, break the last dependency: make migrate_add_blocker() usable before migration_object_init().

The previous commit already removed the use of migrate_get_current() from migrate_add_blocker() itself. Didn't quite do the trick, as there's another one hiding in migration_is_idle().

Quoted 6 lines

The use there isn't actually necessary: when no migration object has been created yet, migration is surely idle. Make migration_is_idle() return true then.

Fixes cda4aa9a5a08777cf13e164c0543bd4888b8adce
Quoted 21 lines
---
 migration/migration.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
migration/migration.c +5 -1
index f6076e5295..609e0df5d0 100644
@@ -1646,7 +1646,11 @@ bool migration_in_postcopy_after_devices(MigrationState *s)
 
 bool migration_is_idle(void)
 {
-    MigrationState *s = migrate_get_current();
+    MigrationState *s = current_migration;
+
+    if (!s) {
+        return true;
+    }
 
     switch (s->state) {
     case MIGRATION_STATUS_NONE:
#15 lore raw

PATCH v2 4/5 vl: Document dependencies hiding in global and compat props

In reply to Markus Armbruster
---
 vl.c | 24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)
vl.c +22 -2
index e4d7ad6b85..8c6d06b01d 100644
@@ -4285,16 +4285,36 @@ int main(int argc, char **argv, char **envp)
     current_machine->maxram_size = maxram_size;
     current_machine->ram_slots = ram_slots;
 
+    /*
+     * Note: uses machine properties such as kernel-irqchip, must run
+     * after machine_set_property().
+     */
     configure_accelerator(current_machine, argv[0]);
 
+    /*
+     * Beware, QOM objects created before this point miss global and
+     * compat properties.
+     *
+     * Global properties get set up by qdev_prop_register_global(),
+     * called from user_register_global_props(), and certain option
+     * desugaring.  Also in CPU feature desugaring (buried in
+     * parse_cpu_model()), which happens below this point, but may
+     * only target the CPU type, which can only be created after
+     * parse_cpu_model() returned the type.
+     *
+     * Machine compat properties: object_set_machine_compat_props().
+     * Accelerator compat props: object_set_accelerator_compat_props(),
+     * called from configure_accelerator().
+     */
+
     if (!qtest_enabled() && machine_class->deprecation_reason) {
         error_report("Machine type '%s' is deprecated: %s",
                      machine_class->name, machine_class->deprecation_reason);
     }
 
     /*
-     * Migration object can only be created after global properties
-     * are applied correctly.
+     * Note: creates a QOM object, must run only after global and
+     * compat properties have been set up.
      */
     migration_object_init();
 
Signature · 1 line
2.17.2
Reviewed-by on v2 4/5
#16 lore raw

Re: [Qemu-devel] [PATCH v2 4/5] vl: Document dependencies hiding in global and compat props

Quoted 1 line · On Mon, 1 Apr 2019 11:08:26 +0200 Markus Armbruster <[email protected]> wrote:
Quoted 47 lines
---
 vl.c | 24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)
vl.c +22 -2
index e4d7ad6b85..8c6d06b01d 100644
@@ -4285,16 +4285,36 @@ int main(int argc, char **argv, char **envp)
     current_machine->maxram_size = maxram_size;
     current_machine->ram_slots = ram_slots;
 
+    /*
+     * Note: uses machine properties such as kernel-irqchip, must run
+     * after machine_set_property().
+     */
     configure_accelerator(current_machine, argv[0]);
 
+    /*
+     * Beware, QOM objects created before this point miss global and
+     * compat properties.
+     *
+     * Global properties get set up by qdev_prop_register_global(),
+     * called from user_register_global_props(), and certain option
+     * desugaring.  Also in CPU feature desugaring (buried in
+     * parse_cpu_model()), which happens below this point, but may
+     * only target the CPU type, which can only be created after
+     * parse_cpu_model() returned the type.
+     *
+     * Machine compat properties: object_set_machine_compat_props().
+     * Accelerator compat props: object_set_accelerator_compat_props(),
+     * called from configure_accelerator().
+     */
+
     if (!qtest_enabled() && machine_class->deprecation_reason) {
         error_report("Machine type '%s' is deprecated: %s",
                      machine_class->name, machine_class->deprecation_reason);
     }
 
     /*
-     * Migration object can only be created after global properties
-     * are applied correctly.
+     * Note: creates a QOM object, must run only after global and
+     * compat properties have been set up.
      */
     migration_object_init();
 
#17 lore raw

Re: [Qemu-devel] [PATCH v2 4/5] vl: Document dependencies hiding in global and compat props

In reply to Markus Armbruster
Quoted 16 lines · * Markus Armbruster ([email protected]) wrote:
---
 vl.c | 24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)
vl.c +3 -0
index e4d7ad6b85..8c6d06b01d 100644
@@ -4285,16 +4285,36 @@ int main(int argc, char **argv, char **envp)
     current_machine->maxram_size = maxram_size;
     current_machine->ram_slots = ram_slots;
 
+    /*
+     * Note: uses machine properties such as kernel-irqchip, must run
+     * after machine_set_property().

vl.c int main(int argc, char **argv, char **envp) · line 4300

Quoted 10 lines
vl.c +8 -0
+     */
     configure_accelerator(current_machine, argv[0]);
 
+    /*
+     * Beware, QOM objects created before this point miss global and
+     * compat properties.
+     *
+     * Global properties get set up by qdev_prop_register_global(),
+     * called from user_register_global_props(), and certain option
+     * desugaring.

Not to stop this patch series; but could we add a way to actually detect when this has happened? Like setting a flag on QOM objects created before this point for example?

Quoted 26 lines
vl.c +11 -2
 Also in CPU feature desugaring (buried in
+     * parse_cpu_model()), which happens below this point, but may
+     * only target the CPU type, which can only be created after
+     * parse_cpu_model() returned the type.
+     *
+     * Machine compat properties: object_set_machine_compat_props().
+     * Accelerator compat props: object_set_accelerator_compat_props(),
+     * called from configure_accelerator().
+     */
+
     if (!qtest_enabled() && machine_class->deprecation_reason) {
         error_report("Machine type '%s' is deprecated: %s",
                      machine_class->name, machine_class->deprecation_reason);
     }
 
     /*
-     * Migration object can only be created after global properties
-     * are applied correctly.
+     * Note: creates a QOM object, must run only after global and
+     * compat properties have been set up.
      */
     migration_object_init();
Signature · 2 lines
-- 
2.17.2

--
Dr. David Alan Gilbert / [email protected] / Manchester, UK

#18 lore raw

Re: [Qemu-devel] [PATCH v2 4/5] vl: Document dependencies hiding in global and compat props

Quoted 20 lines · Dr. David Alan Gilbert
Quoted 19 lines · * Markus Armbruster ([email protected]) wrote:
---
 vl.c | 24 ++++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)
vl.c +4 -0
index e4d7ad6b85..8c6d06b01d 100644
@@ -4285,16 +4285,36 @@ int main(int argc, char **argv, char **envp)
     current_machine->maxram_size = maxram_size;
     current_machine->ram_slots = ram_slots;
 
+    /*
+     * Note: uses machine properties such as kernel-irqchip, must run
+     * after machine_set_property().
+     */
     configure_accelerator(current_machine, argv[0]);
 
Quoted 11 lines

vl.c int main(int argc, char **argv, char **envp)

Quoted 7 lines
vl.c +7 -0
+    /*
+     * Beware, QOM objects created before this point miss global and
+     * compat properties.
+     *
+     * Global properties get set up by qdev_prop_register_global(),
+     * called from user_register_global_props(), and certain option
+     * desugaring.

Not to stop this patch series; but could we add a way to actually detect when this has happened? Like setting a flag on QOM objects created before this point for example?

I started to work on something like this, but had to switch to more urgent work.

#19 lore raw

PATCH v2 5/5 accel: Unbreak accelerator fallback

In reply to Markus Armbruster

When the user specifies a list of accelerators, we pick the first one that initializes successfully. Recent commit 1a3ec8c1564 broke that. Reproducer:

    $ qemu-system-x86_64 --machine accel=xen:tcg
    xencall: error: Could not obtain handle on privileged command interface: No such file or directory
    xen be core: xen be core: can't open xen interface
    can't open xen interface
    qemu-system-x86_64: failed to initialize Xen: Operation not permitted
    qemu-system-x86_64: /home/armbru/work/qemu/qom/object.c:436: object_set_accelerator_compat_props: Assertion `!object_compat_props[0]' failed.

Root cause: we register accelerator compat properties even when the accelerator fails. The failed assertion is object_set_accelerator_compat_props() telling us off. Fix by calling it only for the accelerator that succeeded.

Fixes 1a3ec8c1564f51628cce10d435a2e22559ea29fd
---
 accel/accel.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
accel/accel.c +2 -1
index 8deb475b5d..454fef9d92 100644
@@ -65,8 +65,9 @@ static int accel_init_machine(AccelClass *acc, MachineState *ms)
         ms->accelerator = NULL;
         *(acc->allowed) = false;
         object_unref(OBJECT(accel));
+    } else {
+        object_set_accelerator_compat_props(acc->compat_props);
     }
-    object_set_accelerator_compat_props(acc->compat_props);
     return ret;
 }
 
Signature · 1 line
2.17.2