-incoming defer
v3 cover -incoming defer
From: "Dr. David Alan Gilbert" <[email protected]>
This patchset provides a way of setting options on an incoming migration before the fd/process/socket has been created.
start qemu with -incoming defer
<use the monitor to set whatever you need>
migrate_incoming theuri
v3:
s/pause/defer/ (and associated fixups)
Add ' around the '-incoming defer' in an error
v2:
Create migrate_incoming/migrate-incoming rather than squashing the -u into the existing migrate command.
Dave
Dr. David Alan Gilbert (3):
Add -incoming defer
Add migrate_incoming
Document -incoming options
hmp-commands.hx | 16 ++++++++++++++++ hmp.c | 14 ++++++++++++++ hmp.h | 1 + migration/migration.c | 48 ++++++++++++++++++++++++++++++++++++++++++------ qapi-schema.json | 15 +++++++++++++++ qemu-options.hx | 29 ++++++++++++++++++++++++++--- qmp-commands.hx | 31 ++++++++++++++++++++++++++++++- 7 files changed, 144 insertions(+), 10 deletions(-)
Signature · 1 line
2.1.0
PATCH v3 1/3 Add -incoming defer
From: "Dr. David Alan Gilbert" <[email protected]>
-incoming defer causes qemu to wait for an incoming migration to be specified later. The monitor can be used to set migration capabilities that may affect the incoming connection process.
--- migration/migration.c | 29 +++++++++++++++++++++++------ 1 file changed, 23 insertions(+), 6 deletions(-)
migration/migration.c +23 -6
@@ -49,6 +49,8 @@ enum {static NotifierList migration_state_notifiers =NOTIFIER_LIST_INITIALIZER (migration_state_notifiers); +static bool deferred_incoming; +/* When we add fault tolerance, we could have several migrations at once. For now we don't need to add dynamic creation of migration */
@@ -65,25 +67,40 @@ MigrationState *migrate_get_current(void)return ¤t_migration; } +/* +* Called on -incoming with a defer: uri. +* The migration can be started later after any parameters have been +* changed. +*/ +static void deferred_incoming_migration (Error **errp) +{ +if (deferred_incoming) { +error_setg (errp,"Incoming migration already deferred" ); + } + deferred_incoming = true; +} +void qemu_start_incoming_migration (const char *uri,Error **errp) {const char *p; -if (strstart (uri,"tcp:" , &p)) +if (!strcmp (uri,"defer" )) { +deferred_incoming_migration (errp); + }else if (strstart (uri,"tcp:" , &p)) {tcp_start_incoming_migration (p, errp); #ifdef CONFIG_RDMA - else if (strstart(uri, "rdma:", &p)) + }else if (strstart (uri,"rdma:" , &p)) {rdma_start_incoming_migration (p, errp);#endif #if !defined(WIN32 ) - else if (strstart(uri, "exec:", &p)) + } else if (strstart(uri, "exec:", &p)) {exec_start_incoming_migration (p, errp); - else if (strstart(uri, "unix:", &p)) + }else if (strstart (uri,"unix:" , &p)) {unix_start_incoming_migration (p, errp); - else if (strstart(uri, "fd:", &p)) + }else if (strstart (uri,"fd:" , &p)) {fd_start_incoming_migration (p, errp);#endif - else { + }else {error_setg (errp,"unknown migration protocol: %s" , uri); } }
Signature · 1 line
2.1.0
Re: [Qemu-devel] [PATCH v3 1/3] Add -incoming defer
Quoted 26 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
-incoming defer causes qemu to wait for an incoming migration to be specified later. The monitor can be used to set migration capabilities that may affect the incoming connection process.
Signed-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Juan Quintela <[email protected]>--- migration/migration.c | 29 +++++++++++++++++++++++------ 1 file changed, 23 insertions(+), 6 deletions(-)migration/migration.c +2
-0@@ -49,6 +49,8 @@ enum {static NotifierList migration_state_notifiers =NOTIFIER_LIST_INITIALIZER (migration_state_notifiers); +static bool deferred_incoming; +/* When we add fault tolerance, we could have several migrations at once. For now we don't need to add dynamic creation of migration */
migration/migration.c MigrationState *migrate_get_current(void) · line 71
Quoted 5 lines
migration/migration.c +2
-0@@ -65,25 +67,40 @@ MigrationState *migrate_get_current(void)return ¤t_migration; } +/* +* Called on -incoming with a defer: uri.
The colon in "defer:" is inaccurate, because...
Quoted 7 lines
migration/migration.c +7
-0+* The migration can be started later after any parameters have been +* changed. +*/ +static void deferred_incoming_migration (Error **errp) +{ +if (deferred_incoming) { +error_setg (errp,"Incoming migration already deferred" );
migration/migration.c MigrationState *migrate_get_current(void) · line 87
Quoted 10 lines
migration/migration.c +5
-1+ } + deferred_incoming = true; +} +void qemu_start_incoming_migration (const char *uri,Error **errp) {const char *p; -if (strstart (uri,"tcp:" , &p)) +if (!strcmp (uri,"defer" )) {
... you recognize exactly "defer" here. <pedantic>Which makes it not an URI</pedantic>.
Quoted 14 lines
migration/migration.c +5
-3+deferred_incoming_migration (errp); + }else if (strstart (uri,"tcp:" , &p)) {tcp_start_incoming_migration (p, errp); #ifdef CONFIG_RDMA - else if (strstart(uri, "rdma:", &p)) + }else if (strstart (uri,"rdma:" , &p)) {rdma_start_incoming_migration (p, errp);#endif #if !defined(WIN32 ) - else if (strstart(uri, "exec:", &p)) + } else if (strstart(uri, "exec:", &p)) {exec_start_incoming_migration (p, errp); - else if (strstart(uri, "unix:", &p)) + }else if (strstart (uri,"unix:" , &p)) {
migration/migration.c MigrationState *migrate_get_current(void) · line 106
Quoted 10 lines
migration/migration.c +2
-2unix_start_incoming_migration(p, errp); - else if (strstart(uri, "fd:", &p)) + }else if (strstart (uri,"fd:" , &p)) {fd_start_incoming_migration (p, errp);#endif - else { + }else {error_setg (errp,"unknown migration protocol: %s" , uri); } }
How did you test the new error?
I tried, but ran into this preexisting bug:
$ qemu-system-x86_64 -nodefaults -S -display none -incoming defer -incoming defer
ERROR: invalid runstate transition: 'inmigrate' -> 'inmigrate'
Aborted (core dumped)In my opinion, multiple -incoming should behave like command line options usually do: last one wins silently.
Re: [Qemu-devel] [PATCH v3 1/3] Add -incoming defer
Quoted 31 lines · * Markus Armbruster ([email protected]) wrote:
Quoted 29 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
-incoming defer causes qemu to wait for an incoming migration to be specified later. The monitor can be used to set migration capabilities that may affect the incoming connection process.
Signed-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Juan Quintela <[email protected]>--- migration/migration.c | 29 +++++++++++++++++++++++------ 1 file changed, 23 insertions(+), 6 deletions(-)migration/migration.c +2
-0@@ -49,6 +49,8 @@ enum {static NotifierList migration_state_notifiers =NOTIFIER_LIST_INITIALIZER (migration_state_notifiers); +static bool deferred_incoming; +/* When we add fault tolerance, we could have several migrations at once. For now we don't need to add dynamic creation of migration */ @@ -65,25 +67,40 @@ MigrationState *migrate_get_current(void)return ¤t_migration; }
Quoted 4 lines
migration/migration.c MigrationState *migrate_get_current(void) · line 71
Quoted 2 lines
migration/migration.c +2
-0+/* +* Called on -incoming with a defer: uri. The colon in "defer:" is inaccurate, because...
True, but probably not worth the respin.
Quoted 46 lines
Quoted 7 lines
migration/migration.c +7
-0+* The migration can be started later after any parameters have been +* changed. +*/ +static void deferred_incoming_migration (Error **errp) +{ +if (deferred_incoming) { +error_setg (errp,"Incoming migration already deferred" );migration/migration.c MigrationState *migrate_get_current(void) · line 87
Quoted 10 lines
migration/migration.c +5
-1+ } + deferred_incoming = true; +} +void qemu_start_incoming_migration (const char *uri,Error **errp) {const char *p; -if (strstart (uri,"tcp:" , &p)) +if (!strcmp (uri,"defer" )) {... you recognize exactly "defer" here. <pedantic>Which makes it not an URI</pedantic>.
Quoted 14 lines
migration/migration.c +5
-3+deferred_incoming_migration (errp); + }else if (strstart (uri,"tcp:" , &p)) {tcp_start_incoming_migration (p, errp); #ifdef CONFIG_RDMA - else if (strstart(uri, "rdma:", &p)) + }else if (strstart (uri,"rdma:" , &p)) {rdma_start_incoming_migration (p, errp);#endif #if !defined(WIN32 ) - else if (strstart(uri, "exec:", &p)) + } else if (strstart(uri, "exec:", &p)) {exec_start_incoming_migration (p, errp); - else if (strstart(uri, "unix:", &p)) + }else if (strstart (uri,"unix:" , &p)) {migration/migration.c MigrationState *migrate_get_current(void) · line 106
Quoted 10 lines
migration/migration.c +2
-2unix_start_incoming_migration(p, errp); - else if (strstart(uri, "fd:", &p)) + }else if (strstart (uri,"fd:" , &p)) {fd_start_incoming_migration (p, errp);#endif - else { + }else {error_setg (errp,"unknown migration protocol: %s" , uri); } }
Quoted 7 lines
How did you test the new error?
I tried, but ran into this preexisting bug:
$ qemu-system-x86_64 -nodefaults -S -display none -incoming defer -incoming defer ERROR: invalid runstate transition: 'inmigrate' -> 'inmigrate' Aborted (core dumped)
Fun; yes that's an existing bug that triggers with -incoming tcp::4444 -incoming tcp::4445 I'll think about that separately.
The way I tested the error was like this:
$ bin/qemu-system-x86_64 -nographic -incoming defer QEMU 2.2.50 monitor - type 'help' for more information (qemu) migrate_incoming defer Incoming migration already deferred (qemu)
Quoted 2 lines
In my opinion, multiple -incoming should behave like command line options usually do: last one wins silently.
Either that or error; to me this feels like it should probably error; I'd make the distinction between options that set a value (which works as a last one wins), or options that invoke an action, for options that invoke an action it feels wrong to me to allow incompatible options.
Dave
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Re: [Qemu-devel] [PATCH v3 1/3] Add -incoming defer
Quoted 89 lines · Dr. David Alan Gilbert
Quoted 31 lines · * Markus Armbruster ([email protected]) wrote:
Quoted 29 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
-incoming defer causes qemu to wait for an incoming migration to be specified later. The monitor can be used to set migration capabilities that may affect the incoming connection process.
Signed-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Juan Quintela <[email protected]>--- migration/migration.c | 29 +++++++++++++++++++++++------ 1 file changed, 23 insertions(+), 6 deletions(-)migration/migration.c +2
-0@@ -49,6 +49,8 @@ enum {static NotifierList migration_state_notifiers =NOTIFIER_LIST_INITIALIZER (migration_state_notifiers); +static bool deferred_incoming; +/* When we add fault tolerance, we could have several migrations at once. For now we don't need to add dynamic creation of migration */ @@ -65,25 +67,40 @@ MigrationState *migrate_get_current(void)return ¤t_migration; }Quoted 4 lines
migration/migration.c MigrationState *migrate_get_current(void) · line 71
Quoted 2 lines
migration/migration.c +2
-0+/* +* Called on -incoming with a defer: uri. The colon in "defer:" is inaccurate, because...
True, but probably not worth the respin.
Quoted 50 lines
Quoted 7 lines
migration/migration.c +7
-0+* The migration can be started later after any parameters have been +* changed. +*/ +static void deferred_incoming_migration (Error **errp) +{ +if (deferred_incoming) { +error_setg (errp,"Incoming migration already deferred" );migration/migration.c MigrationState *migrate_get_current(void) · line 87
Quoted 10 lines
migration/migration.c +5
-1+ } + deferred_incoming = true; +} +void qemu_start_incoming_migration (const char *uri,Error **errp) {const char *p; -if (strstart (uri,"tcp:" , &p)) +if (!strcmp (uri,"defer" )) {... you recognize exactly "defer" here. <pedantic>Which makes it not an URI</pedantic>.
Quoted 14 lines
migration/migration.c +5
-3+deferred_incoming_migration (errp); + }else if (strstart (uri,"tcp:" , &p)) {tcp_start_incoming_migration (p, errp); #ifdef CONFIG_RDMA - else if (strstart(uri, "rdma:", &p)) + }else if (strstart (uri,"rdma:" , &p)) {rdma_start_incoming_migration (p, errp);#endif #if !defined(WIN32 ) - else if (strstart(uri, "exec:", &p)) + } else if (strstart(uri, "exec:", &p)) {exec_start_incoming_migration (p, errp); - else if (strstart(uri, "unix:", &p)) + }else if (strstart (uri,"unix:" , &p)) {migration/migration.c MigrationState *migrate_get_current(void) · line 106
Quoted 10 lines
migration/migration.c +2
-2unix_start_incoming_migration(p, errp); - else if (strstart(uri, "fd:", &p)) + }else if (strstart (uri,"fd:" , &p)) {fd_start_incoming_migration (p, errp);#endif - else { + }else {error_setg (errp,"unknown migration protocol: %s" , uri); } }How did you test the new error?
I tried, but ran into this preexisting bug:
Quoted 8 lines
Quoted 4 lines
$ qemu-system-x86_64 -nodefaults -S -display none -incoming defer -incoming defer ERROR: invalid runstate transition: 'inmigrate' -> 'inmigrate' Aborted (core dumped)Fun; yes that's an existing bug that triggers with -incoming tcp::4444 -incoming tcp::4445
I'll think about that separately.
Makes sense.
Quoted 8 lines
The way I tested the error was like this:
$ bin/qemu-system-x86_64 -nographic -incoming defer QEMU 2.2.50 monitor - type 'help' for more information (qemu) migrate_incoming defer Incoming migration already deferred (qemu)
Quoted 8 lines
Quoted 2 lines
In my opinion, multiple -incoming should behave like command line options usually do: last one wins silently.
Either that or error; to me this feels like it should probably error; I'd make the distinction between options that set a value (which works as a last one wins), or options that invoke an action, for options that invoke an action it feels wrong to me to allow incompatible options.
One man's action is another man's value.
There's ample precedence for "last one wins" when command line options conflict. For instance, ls -l conflicts with -C, -m and -x, and POSIX explicitly specifies "last one wins". Rather convenient when you almost always want one option, and use alias to get it without having to type it all the time, but occasionally want a conflicting option, and can just give it despite your alias.
PATCH v3 2/3 Add migrate_incoming
In reply to Dr. David Alan Gilbert (git)From: "Dr. David Alan Gilbert" <[email protected]>
Add migrate_incoming/migrate-incoming to start an incoming migration.
Once a qemu has been started with
-incoming defer
the migration can be started by issuing:
migrate_incoming uri
--- hmp-commands.hx | 16 ++++++++++++++++ hmp.c | 14 ++++++++++++++ hmp.h | 1 + migration/migration.c | 19 +++++++++++++++++++ qapi-schema.json | 15 +++++++++++++++ qmp-commands.hx | 31 ++++++++++++++++++++++++++++++- 6 files changed, 95 insertions(+), 1 deletion(-)
hmp-commands.hx +16 -0
@@ -922,6 +922,22 @@ Cancel the current VM migration. ETEXI { + .name = "migrate_incoming", + .args_type = "uri:s", + .params = "uri", + .help = "Continue an incoming migration from an -incoming defer", + .mhandler.cmd = hmp_migrate_incoming, + }, + +STEXI +@item migrate_incoming @var{uri} +@findex migrate_incoming +Continue an incoming migration using the @var{uri} (that has the same syntax +as the -incoming option). + +ETEXI + + { .name = "migrate_set_cache_size", .args_type = "value:o", .params = "value",
hmp.c +14 -0
@@ -1083,6 +1083,20 @@ void hmp_migrate_cancel(Monitor *mon, const QDict *qdict)qmp_migrate_cancel (NULL ); } +void hmp_migrate_incoming (Monitor *mon,const QDict *qdict) +{ +Error *err =NULL ; +const char *uri =qdict_get_str (qdict,"uri" ); + +qmp_migrate_incoming (uri, &err); + +if (err) { +monitor_printf (mon,"%s\n" ,error_get_pretty (err)); +error_free (err); +return ; + } +} +void hmp_migrate_set_downtime (Monitor *mon,const QDict *qdict) {double value =qdict_get_double (qdict,"value" );
hmp.h +1 -0
@@ -60,6 +60,7 @@ void hmp_snapshot_delete_blkdev_internal(Monitor *mon, const QDict *qdict);void hmp_drive_mirror (Monitor *mon,const QDict *qdict);void hmp_drive_backup (Monitor *mon,const QDict *qdict);void hmp_migrate_cancel (Monitor *mon,const QDict *qdict); +void hmp_migrate_incoming (Monitor *mon,const QDict *qdict);void hmp_migrate_set_downtime (Monitor *mon,const QDict *qdict);void hmp_migrate_set_speed (Monitor *mon,const QDict *qdict);void hmp_migrate_set_capability (Monitor *mon,const QDict *qdict);
migration/migration.c +19 -0
@@ -432,6 +432,25 @@ void migrate_del_blocker(Error *reason) migration_blockers =g_slist_remove (migration_blockers, reason); } +void qmp_migrate_incoming (const char *uri,Error **errp) +{ +Error *local_err =NULL ; + +if (!deferred_incoming) { +error_setg (errp,"'-incoming defer' is required for migrate_incoming" ); +return ; + } + +qemu_start_incoming_migration (uri, &local_err); + +if (local_err) { +error_propagate (errp, local_err); +return ; + } + + deferred_incoming = false; +} +void qmp_migrate (const char *uri,bool has_blk,bool blk,bool has_inc,bool inc,bool has_detach,bool detach,Error **errp)
qapi-schema.json +15 -0
@@ -1738,6 +1738,21 @@ { 'command': 'migrate', 'data': {'uri': 'str', '*blk': 'bool', '*inc': 'bool', '*detach': 'bool' } } +## +# @migrate-incoming +# +# Start an incoming migration, the qemu must have been started +# with -incoming defer +# +# @uri: The Uniform Resource Identifier identifying the source or +# address to listen on +# +# Returns: nothing on success +# +# Since: 2.3 +## +{ 'command': 'migrate-incoming', 'data': {'uri': 'str' } } + # @xen-save-devices-state: # # Save the state of all devices to file. The RAM and the block devices
qmp-commands.hx +30 -1
@@ -661,7 +661,36 @@ Example: <- { "return": {} } EQMP -{ + + { + .name = "migrate-incoming", + .args_type = "uri:s", + .mhandler.cmd_new = qmp_marshal_input_migrate_incoming, + }, + +SQMP +migrate-incoming +---------------- + +Continue an incoming migration + +Arguments: + +- "uri": Source/listening URI (json-string) + +Example: + +-> { "execute": "migrate-incoming", "arguments": { "uri": "tcp::4446" } } +<- { "return": {} } + +Notes: + +(1) QEMU must be started with -incoming defer to allow migrate-incoming to + be used +(2) The uri format is the same as to -incoming + +EQMP + { .name = "migrate-set-cache-size", .args_type = "value:o", .mhandler.cmd_new = qmp_marshal_input_migrate_set_cache_size,
Signature · 1 line
2.1.0
Re: [Qemu-devel] [PATCH v3 2/3] Add migrate_incoming
I'd like Eric's opinion on on encoding configuration tuples as URIs rather than JSON in QMP.
Quoted 59 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
Add migrate_incoming/migrate-incoming to start an incoming migration.
Once a qemu has been started with
-incoming deferthe migration can be started by issuing:
migrate_incominguriSigned-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Eric Blake <[email protected]>--- hmp-commands.hx | 16 ++++++++++++++++ hmp.c | 14 ++++++++++++++ hmp.h | 1 + migration/migration.c | 19 +++++++++++++++++++ qapi-schema.json | 15 +++++++++++++++ qmp-commands.hx | 31 ++++++++++++++++++++++++++++++- 6 files changed, 95 insertions(+), 1 deletion(-)hmp-commands.hx +16
-0@@ -922,6 +922,22 @@ Cancel the current VM migration. ETEXI { + .name = "migrate_incoming", + .args_type = "uri:s", + .params = "uri", + .help = "Continue an incoming migration from an -incoming defer", + .mhandler.cmd = hmp_migrate_incoming, + }, + +STEXI +@item migrate_incoming @var{uri} +@findex migrate_incoming +Continue an incoming migration using the @var{uri} (that has the same syntax +as the -incoming option). + +ETEXI + + { .name = "migrate_set_cache_size", .args_type = "value:o", .params = "value",hmp.c +2
-0@@ -1083,6 +1083,20 @@ void hmp_migrate_cancel(Monitor *mon, const QDict *qdict)qmp_migrate_cancel (NULL ); } +void hmp_migrate_incoming (Monitor *mon,const QDict *qdict) +{
hmp.c void hmp_migrate_cancel(Monitor *mon, const QDict *qdict) · line 1097
Quoted 10 lines
hmp.c +10
-0+Error *err =NULL ; +const char *uri =qdict_get_str (qdict,"uri" ); + +qmp_migrate_incoming (uri, &err); + +if (err) { +monitor_printf (mon,"%s\n" ,error_get_pretty (err)); +error_free (err); +return ; + }
Please replace the conditional by
hmp_handle_error(mon, err);
I'll clean up the other places that should use it.
Quoted 34 lines
hmp.c +2
-0+} +void hmp_migrate_set_downtime (Monitor *mon,const QDict *qdict) {double value =qdict_get_double (qdict,"value" );hmp.h +1
-0@@ -60,6 +60,7 @@ void hmp_snapshot_delete_blkdev_internal(Monitor *mon, const QDict *qdict);void hmp_drive_mirror (Monitor *mon,const QDict *qdict);void hmp_drive_backup (Monitor *mon,const QDict *qdict);void hmp_migrate_cancel (Monitor *mon,const QDict *qdict); +void hmp_migrate_incoming (Monitor *mon,const QDict *qdict);void hmp_migrate_set_downtime (Monitor *mon,const QDict *qdict);void hmp_migrate_set_speed (Monitor *mon,const QDict *qdict);void hmp_migrate_set_capability (Monitor *mon,const QDict *qdict);migration/migration.c +9
-0@@ -432,6 +432,25 @@ void migrate_del_blocker(Error *reason) migration_blockers =g_slist_remove (migration_blockers, reason); } +void qmp_migrate_incoming (const char *uri,Error **errp) +{ +Error *local_err =NULL ; + +if (!deferred_incoming) { +error_setg (errp,"'-incoming defer' is required for migrate_incoming" ); +return ; + } +
migration/migration.c void migrate_del_blocker(Error *reason) · line 453
Quoted 10 lines
migration/migration.c +10
-0+qemu_start_incoming_migration (uri, &local_err); + +if (local_err) { +error_propagate (errp, local_err); +return ; + } + + deferred_incoming = false; +} +
The error message refers to the command as "migrate_incoming", which is wrong for QMP.
Apart from that, the error message is fine when we fail because the user didn't specify -incoming defer. It's unhelpful when we fail because migrate-incoming has already been run.
You could try something like
void qmp_migrate_incoming(const char *uri, Error **errp)
{
Error *local_err = NULL;
if (!deferred_incoming) {
error_setg(errp, "'-incoming defer' is required");
return;
}
if (!runstate_check(RUN_STATE_INMIGRATE)) {
error_setg(errp, "No migration incoming"
return;
}
qemu_start_incoming_migration(uri, &local_err);
if (local_err) {
error_propagate(errp, local_err);
return;
}
}You're welcome to improve my rather laconic error messages.
Quoted 19 lines
migration/migration.c +0
-0void qmp_migrate (const char *uri,bool has_blk,bool blk,bool has_inc,bool inc,bool has_detach,bool detach,Error **errp)qapi-schema.json +8
-0@@ -1738,6 +1738,21 @@ { 'command': 'migrate', 'data': {'uri': 'str', '*blk': 'bool', '*inc': 'bool', '*detach': 'bool' } } +## +# @migrate-incoming +# +# Start an incoming migration, the qemu must have been started +# with -incoming defer +# +# @uri: The Uniform Resource Identifier identifying the source or +# address to listen on
qapi-schema.json line 1758
Quoted 10 lines
qapi-schema.json +7
-0+# +# Returns: nothing on success +# +# Since: 2.3 +## +{ 'command': 'migrate-incoming', 'data': {'uri': 'str' } } + # @xen-save-devices-state: # # Save the state of all devices to file. The RAM and the block devices
Eric, what's your take on this?
The general rule in QMP is "no ad hoc encoding of tuples in strings, use JSON objects instead".
Quoted 42 lines
qmp-commands.hx +30
-1@@ -661,7 +661,36 @@ Example: <- { "return": {} } EQMP -{ + + { + .name = "migrate-incoming", + .args_type = "uri:s", + .mhandler.cmd_new = qmp_marshal_input_migrate_incoming, + }, + +SQMP +migrate-incoming +---------------- + +Continue an incoming migration + +Arguments: + +- "uri": Source/listening URI (json-string) + +Example: + +-> { "execute": "migrate-incoming", "arguments": { "uri": "tcp::4446" } } +<- { "return": {} } + +Notes: + +(1) QEMU must be started with -incoming defer to allow migrate-incoming to + be used +(2) The uri format is the same as to -incoming + +EQMP + { .name = "migrate-set-cache-size", .args_type = "value:o", .mhandler.cmd_new = qmp_marshal_input_migrate_set_cache_size,
Re: [Qemu-devel] [PATCH v3 2/3] Add migrate_incoming
Quoted 2 lines · * Markus Armbruster ([email protected]) wrote:
I'd like Eric's opinion on on encoding configuration tuples as URIs rather than JSON in QMP.
Old discussion; he's already R-b the previous version of this patch where the only difference was pause/defer.
Dave
Quoted 225 lines
Quoted 59 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
Add migrate_incoming/migrate-incoming to start an incoming migration.
Once a qemu has been started with
-incoming deferthe migration can be started by issuing:
migrate_incominguriSigned-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Eric Blake <[email protected]>--- hmp-commands.hx | 16 ++++++++++++++++ hmp.c | 14 ++++++++++++++ hmp.h | 1 + migration/migration.c | 19 +++++++++++++++++++ qapi-schema.json | 15 +++++++++++++++ qmp-commands.hx | 31 ++++++++++++++++++++++++++++++- 6 files changed, 95 insertions(+), 1 deletion(-)hmp-commands.hx +16
-0@@ -922,6 +922,22 @@ Cancel the current VM migration. ETEXI { + .name = "migrate_incoming", + .args_type = "uri:s", + .params = "uri", + .help = "Continue an incoming migration from an -incoming defer", + .mhandler.cmd = hmp_migrate_incoming, + }, + +STEXI +@item migrate_incoming @var{uri} +@findex migrate_incoming +Continue an incoming migration using the @var{uri} (that has the same syntax +as the -incoming option). + +ETEXI + + { .name = "migrate_set_cache_size", .args_type = "value:o", .params = "value",hmp.c +2
-0@@ -1083,6 +1083,20 @@ void hmp_migrate_cancel(Monitor *mon, const QDict *qdict)qmp_migrate_cancel (NULL ); } +void hmp_migrate_incoming (Monitor *mon,const QDict *qdict) +{hmp.c void hmp_migrate_cancel(Monitor *mon, const QDict *qdict) · line 1097
Quoted 10 lines
hmp.c +10
-0+Error *err =NULL ; +const char *uri =qdict_get_str (qdict,"uri" ); + +qmp_migrate_incoming (uri, &err); + +if (err) { +monitor_printf (mon,"%s\n" ,error_get_pretty (err)); +error_free (err); +return ; + }Please replace the conditional by
hmp_handle_error(mon, err);I'll clean up the other places that should use it.
Quoted 34 lines
hmp.c +2
-0+} +void hmp_migrate_set_downtime (Monitor *mon,const QDict *qdict) {double value =qdict_get_double (qdict,"value" );hmp.h +1
-0@@ -60,6 +60,7 @@ void hmp_snapshot_delete_blkdev_internal(Monitor *mon, const QDict *qdict);void hmp_drive_mirror (Monitor *mon,const QDict *qdict);void hmp_drive_backup (Monitor *mon,const QDict *qdict);void hmp_migrate_cancel (Monitor *mon,const QDict *qdict); +void hmp_migrate_incoming (Monitor *mon,const QDict *qdict);void hmp_migrate_set_downtime (Monitor *mon,const QDict *qdict);void hmp_migrate_set_speed (Monitor *mon,const QDict *qdict);void hmp_migrate_set_capability (Monitor *mon,const QDict *qdict);migration/migration.c +9
-0@@ -432,6 +432,25 @@ void migrate_del_blocker(Error *reason) migration_blockers =g_slist_remove (migration_blockers, reason); } +void qmp_migrate_incoming (const char *uri,Error **errp) +{ +Error *local_err =NULL ; + +if (!deferred_incoming) { +error_setg (errp,"'-incoming defer' is required for migrate_incoming" ); +return ; + } +migration/migration.c void migrate_del_blocker(Error *reason) · line 453
Quoted 10 lines
migration/migration.c +10
-0+qemu_start_incoming_migration (uri, &local_err); + +if (local_err) { +error_propagate (errp, local_err); +return ; + } + + deferred_incoming = false; +} +The error message refers to the command as "
migrate_incoming", which is wrong for QMP.Apart from that, the error message is fine when we fail because the user didn't specify -incoming defer. It's unhelpful when we fail because migrate-incoming has already been run.
You could try something like
void qmp_migrate_incoming(const char *uri, Error **errp) { Error *local_err = NULL; if (!deferred_incoming) { error_setg(errp, "'-incoming defer' is required"); return; } if (!runstate_check(RUN_STATE_INMIGRATE)) { error_setg(errp, "No migration incoming" return; } qemu_start_incoming_migration(uri, &local_err); if (local_err) { error_propagate(errp, local_err); return; } }You're welcome to improve my rather laconic error messages.
Quoted 19 lines
migration/migration.c +0
-0void qmp_migrate (const char *uri,bool has_blk,bool blk,bool has_inc,bool inc,bool has_detach,bool detach,Error **errp)qapi-schema.json +8
-0@@ -1738,6 +1738,21 @@ { 'command': 'migrate', 'data': {'uri': 'str', '*blk': 'bool', '*inc': 'bool', '*detach': 'bool' } } +## +# @migrate-incoming +# +# Start an incoming migration, the qemu must have been started +# with -incoming defer +# +# @uri: The Uniform Resource Identifier identifying the source or +# address to listen onqapi-schema.json line 1758
Quoted 10 lines
qapi-schema.json +7
-0+# +# Returns: nothing on success +# +# Since: 2.3 +## +{ 'command': 'migrate-incoming', 'data': {'uri': 'str' } } + # @xen-save-devices-state: # # Save the state of all devices to file. The RAM and the block devicesEric, what's your take on this?
The general rule in QMP is "no ad hoc encoding of tuples in strings, use JSON objects instead".
Quoted 34 lines
qmp-commands.hx +25
-1@@ -661,7 +661,36 @@ Example: <- { "return": {} } EQMP -{ + + { + .name = "migrate-incoming", + .args_type = "uri:s", + .mhandler.cmd_new = qmp_marshal_input_migrate_incoming, + }, + +SQMP +migrate-incoming +---------------- + +Continue an incoming migration + +Arguments: + +- "uri": Source/listening URI (json-string) + +Example: + +-> { "execute": "migrate-incoming", "arguments": { "uri": "tcp::4446" } } +<- { "return": {} } + +Notes: + +(1) QEMU must be started with -incoming defer to allow migrate-incoming to
Quoted 8 lines
qmp-commands.hx Example: · line 696
Quoted 8 lines
qmp-commands.hx +5
-0+ be used +(2) The uri format is the same as to -incoming + +EQMP + { .name = "migrate-set-cache-size", .args_type = "value:o", .mhandler.cmd_new = qmp_marshal_input_migrate_set_cache_size,
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Re: [Qemu-devel] [PATCH v3 2/3] Add migrate_incoming
Quoted 6 lines · Dr. David Alan Gilbert
Quoted 2 lines · * Markus Armbruster ([email protected]) wrote:
I'd like Eric's opinion on on encoding configuration tuples as URIs rather than JSON in QMP.
Old discussion; he's already R-b the previous version of this patch where the only difference was pause/defer.
i will send the pull request for this (actually previous version was on my list already). Please update Markus comment as a separate serie.
Thanks, Juan.
Quoted 238 lines
Dave
Quoted 225 lines
Quoted 59 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
Add migrate_incoming/migrate-incoming to start an incoming migration.
Once a qemu has been started with
-incoming deferthe migration can be started by issuing:
migrate_incominguriSigned-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Eric Blake <[email protected]>--- hmp-commands.hx | 16 ++++++++++++++++ hmp.c | 14 ++++++++++++++ hmp.h | 1 + migration/migration.c | 19 +++++++++++++++++++ qapi-schema.json | 15 +++++++++++++++ qmp-commands.hx | 31 ++++++++++++++++++++++++++++++- 6 files changed, 95 insertions(+), 1 deletion(-)hmp-commands.hx +16
-0@@ -922,6 +922,22 @@ Cancel the current VM migration. ETEXI { + .name = "migrate_incoming", + .args_type = "uri:s", + .params = "uri", + .help = "Continue an incoming migration from an -incoming defer", + .mhandler.cmd = hmp_migrate_incoming, + }, + +STEXI +@item migrate_incoming @var{uri} +@findex migrate_incoming +Continue an incoming migration using the @var{uri} (that has the same syntax +as the -incoming option). + +ETEXI + + { .name = "migrate_set_cache_size", .args_type = "value:o", .params = "value",hmp.c +2
-0@@ -1083,6 +1083,20 @@ void hmp_migrate_cancel(Monitor *mon, const QDict *qdict)qmp_migrate_cancel (NULL ); } +void hmp_migrate_incoming (Monitor *mon,const QDict *qdict) +{hmp.c void hmp_migrate_cancel(Monitor *mon, const QDict *qdict) · line 1097
Quoted 10 lines
hmp.c +10
-0+Error *err =NULL ; +const char *uri =qdict_get_str (qdict,"uri" ); + +qmp_migrate_incoming (uri, &err); + +if (err) { +monitor_printf (mon,"%s\n" ,error_get_pretty (err)); +error_free (err); +return ; + }Please replace the conditional by
hmp_handle_error(mon, err);I'll clean up the other places that should use it.
Quoted 34 lines
hmp.c +2
-0+} +void hmp_migrate_set_downtime (Monitor *mon,const QDict *qdict) {double value =qdict_get_double (qdict,"value" );hmp.h +1
-0@@ -60,6 +60,7 @@ void hmp_snapshot_delete_blkdev_internal(Monitor *mon, const QDict *qdict);void hmp_drive_mirror (Monitor *mon,const QDict *qdict);void hmp_drive_backup (Monitor *mon,const QDict *qdict);void hmp_migrate_cancel (Monitor *mon,const QDict *qdict); +void hmp_migrate_incoming (Monitor *mon,const QDict *qdict);void hmp_migrate_set_downtime (Monitor *mon,const QDict *qdict);void hmp_migrate_set_speed (Monitor *mon,const QDict *qdict);void hmp_migrate_set_capability (Monitor *mon,const QDict *qdict);migration/migration.c +9
-0@@ -432,6 +432,25 @@ void migrate_del_blocker(Error *reason) migration_blockers =g_slist_remove (migration_blockers, reason); } +void qmp_migrate_incoming (const char *uri,Error **errp) +{ +Error *local_err =NULL ; + +if (!deferred_incoming) { +error_setg (errp,"'-incoming defer' is required for migrate_incoming" ); +return ; + } +migration/migration.c void migrate_del_blocker(Error *reason) · line 453
Quoted 10 lines
migration/migration.c +10
-0+qemu_start_incoming_migration (uri, &local_err); + +if (local_err) { +error_propagate (errp, local_err); +return ; + } + + deferred_incoming = false; +} +The error message refers to the command as "
migrate_incoming", which is wrong for QMP.Apart from that, the error message is fine when we fail because the user didn't specify -incoming defer. It's unhelpful when we fail because migrate-incoming has already been run.
You could try something like
void qmp_migrate_incoming(const char *uri, Error **errp) { Error *local_err = NULL; if (!deferred_incoming) { error_setg(errp, "'-incoming defer' is required"); return; } if (!runstate_check(RUN_STATE_INMIGRATE)) { error_setg(errp, "No migration incoming" return; } qemu_start_incoming_migration(uri, &local_err); if (local_err) { error_propagate(errp, local_err); return; } }You're welcome to improve my rather laconic error messages.
Quoted 19 lines
migration/migration.c +0
-0void qmp_migrate (const char *uri,bool has_blk,bool blk,bool has_inc,bool inc,bool has_detach,bool detach,Error **errp)qapi-schema.json +8
-0@@ -1738,6 +1738,21 @@ { 'command': 'migrate', 'data': {'uri': 'str', '*blk': 'bool', '*inc': 'bool', '*detach': 'bool' } } +## +# @migrate-incoming +# +# Start an incoming migration, the qemu must have been started +# with -incoming defer +# +# @uri: The Uniform Resource Identifier identifying the source or +# address to listen onqapi-schema.json line 1758
Quoted 10 lines
qapi-schema.json +7
-0+# +# Returns: nothing on success +# +# Since: 2.3 +## +{ 'command': 'migrate-incoming', 'data': {'uri': 'str' } } + # @xen-save-devices-state: # # Save the state of all devices to file. The RAM and the block devicesEric, what's your take on this?
The general rule in QMP is "no ad hoc encoding of tuples in strings, use JSON objects instead".
Quoted 34 lines
qmp-commands.hx +25
-1@@ -661,7 +661,36 @@ Example: <- { "return": {} } EQMP -{ + + { + .name = "migrate-incoming", + .args_type = "uri:s", + .mhandler.cmd_new = qmp_marshal_input_migrate_incoming, + }, + +SQMP +migrate-incoming +---------------- + +Continue an incoming migration + +Arguments: + +- "uri": Source/listening URI (json-string) + +Example: + +-> { "execute": "migrate-incoming", "arguments": { "uri": "tcp::4446" } } +<- { "return": {} } + +Notes: + +(1) QEMU must be started with -incoming defer to allow migrate-incoming toQuoted 8 lines
qmp-commands.hx Example: · line 696
Quoted 8 lines
qmp-commands.hx +5
-0+ be used +(2) The uri format is the same as to -incoming + +EQMP + { .name = "migrate-set-cache-size", .args_type = "value:o", .mhandler.cmd_new = qmp_marshal_input_migrate_set_cache_size,--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Re: [Qemu-devel] [PATCH v3 2/3] Add migrate_incoming
In reply to Markus ArmbrusterQuoted 12 lines · Markus Armbruster
I'd like Eric's opinion on on encoding configuration tuples as URIs rather than JSON in QMP.
qapi-schema.json line 1758
Quoted 5 lines
qapi-schema.json +2
-0+{ 'command': 'migrate-incoming', 'data': {'uri': 'str' } } + # @xen-save-devices-state: # # Save the state of all devices to file. The RAM and the block devicesEric, what's your take on this?
Quoted 2 lines
The general rule in QMP is "no ad hoc encoding of tuples in strings, use JSON objects instead".
Yes, it would be nice to be type-safe, and have a QAPI union with a discriminator of different URI types along with appropriate accompanying data. But it is also new code (both in qemu to parse it, and in libvirt to generate it), and we already have existing code that knows how to generate the string-encoded tuple for the -incoming command line argument that can be reused as-is with the new QMP command as proposed.
So even though it is less safe, I'm okay with this particular command
using an overloaded string argument.
Down the road, if we WANT to add type-safety, we can make this command take an anonymous union, where a 'str' value is the compact old style, and where a dictionary value is the new discriminated union style. So we aren't completely locked into non-type-safe string forever, but we probably won't add a type-safe variant until (if) we ever add a new migration format that just looks too ugly as a URI based on the parameters it requires.
Signature · 2 lines
Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
Re: [Qemu-devel] [PATCH v3 2/3] Add migrate_incoming
Quoted 25 lines · Eric Blake
Quoted 12 lines · Markus Armbruster
I'd like Eric's opinion on on encoding configuration tuples as URIs rather than JSON in QMP.
qapi-schema.json line 1758
Quoted 5 lines
qapi-schema.json +2
-0+{ 'command': 'migrate-incoming', 'data': {'uri': 'str' } } + # @xen-save-devices-state: # # Save the state of all devices to file. The RAM and the block devicesEric, what's your take on this?
Quoted 2 lines
The general rule in QMP is "no ad hoc encoding of tuples in strings, use JSON objects instead".
Yes, it would be nice to be type-safe, and have a QAPI union with a discriminator of different URI types along with appropriate accompanying data. But it is also new code (both in qemu to parse it, and in libvirt to generate it), and we already have existing code that knows how to generate the string-encoded tuple for the -incoming command line argument that can be reused as-is with the new QMP command as proposed.
So even though it is less safe, I'm okay with this particular command
using an overloaded string argument.
Quoted 7 lines
Down the road, if we WANT to add type-safety, we can make this command take an anonymous union, where a 'str' value is the compact old style, and where a dictionary value is the new discriminated union style. So we aren't completely locked into non-type-safe string forever, but we probably won't add a type-safe variant until (if) we ever add a new migration format that just looks too ugly as a URI based on the parameters it requires.
I can accept this, but I want the rationale in a comment, so this exception to the QMP rule serves less well as bad example.
Re: [Qemu-devel] [PATCH v3 2/3] Add migrate_incoming
In reply to Markus ArmbrusterQuoted 2 lines · * Markus Armbruster ([email protected]) wrote:
I'd like Eric's opinion on on encoding configuration tuples as URIs rather than JSON in QMP.
I've just posted a series which should address all the issues in here; but note that:
Quoted 50 lines
Quoted 9 lines
migration/migration.c +9
-0+void qmp_migrate_incoming (const char *uri,Error **errp) +{ +Error *local_err =NULL ; + +if (!deferred_incoming) { +error_setg (errp,"'-incoming defer' is required for migrate_incoming" ); +return ; + } +migration/migration.c void migrate_del_blocker(Error *reason) · line 453
Quoted 10 lines
migration/migration.c +10
-0+qemu_start_incoming_migration (uri, &local_err); + +if (local_err) { +error_propagate (errp, local_err); +return ; + } + + deferred_incoming = false; +} +The error message refers to the command as "
migrate_incoming", which is wrong for QMP.Apart from that, the error message is fine when we fail because the user didn't specify -incoming defer. It's unhelpful when we fail because migrate-incoming has already been run.
You could try something like
void qmp_migrate_incoming(const char *uri, Error **errp) { Error *local_err = NULL; if (!deferred_incoming) { error_setg(errp, "'-incoming defer' is required"); return; } if (!runstate_check(RUN_STATE_INMIGRATE)) { error_setg(errp, "No migration incoming" return; } qemu_start_incoming_migration(uri, &local_err); if (local_err) { error_propagate(errp, local_err); return; } }
Quoted 1 line
You're welcome to improve my rather laconic error messages.
doesn't work; because qemu_start_incoming_migration returns immediately, and so:
migrate_incoming tcp::4444migrate_incoming tcp::4444
fails trying to bind the port twice. See the new patch series for my way of fixing it.
Dave
Quoted 70 lines
Quoted 20 lines
void qmp_migrate(const char *uri, bool
has_blk, bool blk,
boolhas_inc, bool inc, boolhas_detach, bool detach, Error **errp)qapi-schema.json +8
-0@@ -1738,6 +1738,21 @@ { 'command': 'migrate', 'data': {'uri': 'str', '*blk': 'bool', '*inc': 'bool', '*detach': 'bool' } } +## +# @migrate-incoming +# +# Start an incoming migration, the qemu must have been started +# with -incoming defer +# +# @uri: The Uniform Resource Identifier identifying the source or +# address to listen onqapi-schema.json line 1758
Quoted 10 lines
qapi-schema.json +7
-0+# +# Returns: nothing on success +# +# Since: 2.3 +## +{ 'command': 'migrate-incoming', 'data': {'uri': 'str' } } + # @xen-save-devices-state: # # Save the state of all devices to file. The RAM and the block devicesEric, what's your take on this?
The general rule in QMP is "no ad hoc encoding of tuples in strings, use JSON objects instead".
Quoted 34 lines
qmp-commands.hx +25
-1@@ -661,7 +661,36 @@ Example: <- { "return": {} } EQMP -{ + + { + .name = "migrate-incoming", + .args_type = "uri:s", + .mhandler.cmd_new = qmp_marshal_input_migrate_incoming, + }, + +SQMP +migrate-incoming +---------------- + +Continue an incoming migration + +Arguments: + +- "uri": Source/listening URI (json-string) + +Example: + +-> { "execute": "migrate-incoming", "arguments": { "uri": "tcp::4446" } } +<- { "return": {} } + +Notes: + +(1) QEMU must be started with -incoming defer to allow migrate-incoming to
Quoted 8 lines
qmp-commands.hx Example: · line 696
Quoted 8 lines
qmp-commands.hx +5
-0+ be used +(2) The uri format is the same as to -incoming + +EQMP + { .name = "migrate-set-cache-size", .args_type = "value:o", .mhandler.cmd_new = qmp_marshal_input_migrate_set_cache_size,
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
PATCH v3 3/3 Document -incoming options
In reply to Dr. David Alan Gilbert (git)From: "Dr. David Alan Gilbert" <[email protected]>
Document the various URI formats for -incoming, the previous manpage and help text was wrong (out of date?)
--- qemu-options.hx | 29 ++++++++++++++++++++++++++--- 1 file changed, 26 insertions(+), 3 deletions(-)
qemu-options.hx +26 -3
@@ -3169,12 +3169,35 @@ Set TB size. ETEXI DEF("incoming", HAS_ARG, QEMU_OPTION_incoming, \ - "-incoming p prepare for incoming migration, listen on port p\n", + "-incoming uri prepare for incoming migration, specifying source:\n" \ + " exec:command Execute 'command' use the stdout as\n" \ + " the migration stream\n" \ + " fd:num listen on the given fd\n" \ + " defer wait for the URI to be specified by\n" \ + " the monitor (migrate_incoming)\n" \ + " rdma:addr:port Listen on RDMA port on given address\n" \ + " tcp:addr:port listen on TCP port (optional address)\n" \ + " unix:path listen on the UNIX socket 'path'\n", \ QEMU_ARCH_ALL) STEXI -@item -incoming @var{port} +@item -incoming @var{uri} @findex -incoming -Prepare for incoming migration, listen on @var{port}. +Prepare for incoming migration, specifying the source of the migration stream +@table @option +@item exec:@var{command} +Execute 'command' and use the stdout as the migration stream. +@item fd:@var{num} +listen on the given fd +@item defer +wait for the URI to be specified by the monitor (migrate_incoming) +@item rdma:@var{addr}:@var{port} +Listen on RDMA port on given address +@item tcp:@var{addr}:@var{port}[,ipv4][,ipv6][,to=to] +Listen on TCP port @var{port} (optional @var{addr} to specify address to listen on). +The options ,ipv4, ipv6 and ,to are used in the same manner as chardev TCP options. +@item unix:@var{path} +listen on the UNIX socket @var{path} +@end table ETEXI DEF("nodefaults", 0, QEMU_OPTION_nodefaults, \
Signature · 1 line
2.1.0
Re: [Qemu-devel] [PATCH v3 3/3] Document -incoming options
Quoted 4 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
Document the various URI formats for -incoming, the previous manpage and help text was wrong (out of date?)
Thanks a lot for updating the docs.
Quoted 25 lines
Signed-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Juan Quintela <[email protected]>--- qemu-options.hx | 29 ++++++++++++++++++++++++++--- 1 file changed, 26 insertions(+), 3 deletions(-)qemu-options.hx +9
-1@@ -3169,12 +3169,35 @@ Set TB size. ETEXI DEF("incoming", HAS_ARG, QEMU_OPTION_incoming, \ - "-incoming p prepare for incoming migration, listen on port p\n", + "-incoming uri prepare for incoming migration, specifying source:\n" \ + " exec:command Execute 'command' use the stdout as\n" \ + " the migration stream\n" \ + " fd:num listen on the given fd\n" \ + " defer wait for the URI to be specified by\n" \ + " the monitor (migrate_incoming)\n" \ + " rdma:addr:port Listen on RDMA port on given address\n" \ + " tcp:addr:port listen on TCP port (optional address)\n" \ + " unix:path listen on the UNIX socket 'path'\n", \
qemu-options.hx Set TB size. · line 3188
Quoted 10 lines
qemu-options.hx +5
-2QEMU_ARCH_ALL) STEXI -@item -incoming @var{port} +@item -incoming @var{uri} @findex -incoming -Prepare for incoming migration, listen on @var{port}. +Prepare for incoming migration, specifying the source of the migration stream +@table @option +@item exec:@var{command} +Execute 'command' and use the stdout as the migration stream.
Suggest:
Execute 'command' and use its standard output as migration stream source.
qemu-options.hx Set TB size. · line 3190
Quoted 2 lines
qemu-options.hx +2
-0+@item fd:@var{num} +listen on the given fd
Well, we're not listening in usual the sense of listen(2). Suggest
Use file descriptor @var{num}.
qemu-options.hx Set TB size. · line 3192
Quoted 2 lines
qemu-options.hx +2
-0+@item defer +wait for the URI to be specified by the monitor (migrate_incoming)
"Wait", since the other items start with a capital letter.
qemu-options.hx Set TB size. · line 3194
Quoted 2 lines
qemu-options.hx +2
-0+@item rdma:@var{addr}:@var{port} +Listen on RDMA port on given address
Let's call the thing between the colons "host" instead of "äddr". The address consists of host and port.
rdma_start_incoming_migration() passes everything after rdma: to qemu_rdma_data_init(), which parses it with inet_parse(), then uses only host and port. The other members of InetSocketAddress are silently ignored. Wonderful.
qemu_rdma_data_init() errors out if you omit host.
inet_parse() accepts one of
:PORT
[V6ADDR]:PORT
V4ADDR:PORT
HOST:PORTfollowed by an options string. The options string isn't really parsed, instead substrings matching one of these patterns are recognized:
,to=PORT
,ipv4
,ipv6Note: you can specify ipv4,ipv6 or combine V6ADDR with ipv4, or V4ADDR with ipv6 to get both flags set in InetSocketAddress.
qemu-options.hx Set TB size. · line 3195
Quoted 1 line
qemu-options.hx +1
-0+@item tcp:@var{addr}:@var{port}[,ipv4][,ipv6][,to=to]
[@var{host}] instead of @var{addr}, and [,to=@var{port}] instead of [,to=to], please.
qemu-options.hx Set TB size. · line 3197
Quoted 2 lines
qemu-options.hx +2
-0+Listen on TCP port @var{port} (optional @var{addr} to specify address to listen on). +The options ,ipv4, ipv6 and ,to are used in the same manner as chardev TCP options.
The reference to chardev options makes sense, but the reader needs to make the connection from "chardev TCP options" to "-chardev socket" himself. Maybe "in the same manner as in -chardev socket"?
qemu-options.hx Set TB size. · line 3203
Quoted 6 lines
qemu-options.hx +3
-0+@item unix:@var{path} +listen on the UNIX socket @var{path} +@end table ETEXI DEF("nodefaults", 0, QEMU_OPTION_nodefaults, \
Don't get fooled by my many comments, this series is good stuff!
Re: [Qemu-devel] [PATCH v3 3/3] Document -incoming options
Quoted 47 lines · * Markus Armbruster ([email protected]) wrote:
Quoted 4 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
Document the various URI formats for -incoming, the previous manpage and help text was wrong (out of date?)
Thanks a lot for updating the docs.
Quoted 25 lines
Signed-off-by Dr. David Alan Gilbert <[email protected]>Reviewed-by Juan Quintela <[email protected]>--- qemu-options.hx | 29 ++++++++++++++++++++++++++--- 1 file changed, 26 insertions(+), 3 deletions(-)qemu-options.hx +9
-1@@ -3169,12 +3169,35 @@ Set TB size. ETEXI DEF("incoming", HAS_ARG, QEMU_OPTION_incoming, \ - "-incoming p prepare for incoming migration, listen on port p\n", + "-incoming uri prepare for incoming migration, specifying source:\n" \ + " exec:command Execute 'command' use the stdout as\n" \ + " the migration stream\n" \ + " fd:num listen on the given fd\n" \ + " defer wait for the URI to be specified by\n" \ + " the monitor (migrate_incoming)\n" \ + " rdma:addr:port Listen on RDMA port on given address\n" \ + " tcp:addr:port listen on TCP port (optional address)\n" \ + " unix:path listen on the UNIX socket 'path'\n", \qemu-options.hx Set TB size. · line 3188
Quoted 10 lines
qemu-options.hx +5
-2QEMU_ARCH_ALL) STEXI -@item -incoming @var{port} +@item -incoming @var{uri} @findex -incoming -Prepare for incoming migration, listen on @var{port}. +Prepare for incoming migration, specifying the source of the migration stream +@table @option +@item exec:@var{command} +Execute 'command' and use the stdout as the migration stream.Suggest:
Quoted 2 lines
Execute 'command' and use its standard output as migration stream source.
Yes OK, can do that.
Quoted 6 lines
qemu-options.hx Set TB size. · line 3190
Quoted 2 lines
qemu-options.hx +2
-0+@item fd:@var{num} +listen on the given fdWell, we're not listening in usual the sense of listen(2). Suggest
Use file descriptor @var{num}.
OK.
Quoted 10 lines
qemu-options.hx Set TB size. · line 3192
Quoted 2 lines
qemu-options.hx +2
-0+@item defer +wait for the URI to be specified by the monitor (migrate_incoming)"Wait", since the other items start with a capital letter.
qemu-options.hx Set TB size. · line 3194
Quoted 2 lines
qemu-options.hx +2
-0+@item rdma:@var{addr}:@var{port} +Listen on RDMA port on given addressLet's call the thing between the colons "host" instead of "addr". The address consists of host and port.
Yes I thought about doing that, but it really isn't a host, it's the identifier for the network cards you're going to listen on; saying 'host' suggests thats that you need to specify the source host of the destination which is certainly wrong.
Quoted 4 lines
rdma_start_incoming_migration()passes everything after rdma: toqemu_rdma_data_init(), which parses it withinet_parse(), then uses only host and port. The other members of InetSocketAddress are silently ignored. Wonderful.
Yes, there are also some weird dependencies in the RDMA code on IPv6 availability I've not quite figured out yet.
Quoted 16 lines
qemu_rdma_data_init()errors out if you omit host.
inet_parse()accepts one of:PORT [V6ADDR]:PORT V4ADDR:PORT HOST:PORTfollowed by an options string. The options string isn't really parsed, instead substrings matching one of these patterns are recognized:
,to=PORT ,ipv4 ,ipv6
Quoted 7 lines
Note: you can specify ipv4,ipv6 or combine V6ADDR with ipv4, or V4ADDR with ipv6 to get both flags set in InetSocketAddress.
qemu-options.hx Set TB size. · line 3195
Quoted 1 line
qemu-options.hx +1
-0+@item tcp:@var{addr}:@var{port}[,ipv4][,ipv6][,to=to][@var{host}] instead of @var{addr}, and [,to=@var{port}] instead of [,to=to], please.
OK.
Quoted 6 lines
qemu-options.hx Set TB size. · line 3197
Quoted 2 lines
qemu-options.hx +2
-0+Listen on TCP port @var{port} (optional @var{addr} to specify address to listen on). +The options ,ipv4, ipv6 and ,to are used in the same manner as chardev TCP options.The reference to chardev options makes sense, but the reader needs to make the connection from "chardev TCP options" to "-chardev socket" himself. Maybe "in the same manner as in -chardev socket"?
Yep OK.
Quoted 8 lines
qemu-options.hx Set TB size. · line 3203
Quoted 6 lines
qemu-options.hx +3
-0+@item unix:@var{path} +listen on the UNIX socket @var{path} +@end table ETEXI DEF("nodefaults", 0, QEMU_OPTION_nodefaults, \Don't get fooled by my many comments, this series is good stuff!
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Quoted 24 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
This patchset provides a way of setting options on an incoming migration before the fd/process/socket has been created.
start qemu with -incoming defer
<use the monitor to set whatever you need>
migrate_incomingtheuriv3:
s/pause/defer/ (and associated fixups)
Add ' around the '-incoming defer' in an errorv2:
Create migrate_incoming/migrate-incoming rather than squashing the -u into the existing migrate command.Dave
Dr. David Alan Gilbert (3):
Add -incoming defer
Addmigrate_incoming
Document -incoming options
Quoted 8 lines
hmp-commands.hx | 16 ++++++++++++++++ hmp.c | 14 ++++++++++++++ hmp.h | 1 + migration/migration.c | 48 ++++++++++++++++++++++++++++++++++++++++++------ qapi-schema.json | 15 +++++++++++++++ qemu-options.hx | 29 ++++++++++++++++++++++++++--- qmp-commands.hx | 31 ++++++++++++++++++++++++++++++- 7 files changed, 144 insertions(+), 10 deletions(-)
Update qemu-doc.texi as well?
Quoted 3 lines
Signature · 1 line
2.1.0
Quoted 34 lines · * Michael S. Tsirkin ([email protected]) wrote:
Quoted 24 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
This patchset provides a way of setting options on an incoming migration before the fd/process/socket has been created.
start qemu with -incoming defer
<use the monitor to set whatever you need>
migrate_incomingtheuriv3:
s/pause/defer/ (and associated fixups)
Add ' around the '-incoming defer' in an errorv2:
Create migrate_incoming/migrate-incoming rather than squashing the -u into the existing migrate command.Dave
Dr. David Alan Gilbert (3):
Add -incoming defer
Addmigrate_incoming
Document -incoming optionsQuoted 8 lines
hmp-commands.hx | 16 ++++++++++++++++ hmp.c | 14 ++++++++++++++ hmp.h | 1 + migration/migration.c | 48 ++++++++++++++++++++++++++++++++++++++++++------ qapi-schema.json | 15 +++++++++++++++ qemu-options.hx | 29 ++++++++++++++++++++++++++--- qmp-commands.hx | 31 ++++++++++++++++++++++++++++++- 7 files changed, 144 insertions(+), 10 deletions(-)
Quoted 1 line
Update qemu-doc.texi as well?
Well the qemu-doc.html seems to have gained the migrate_incoming command and proper definitions of -incoming magically without me having to look at a texi file - I think that description is parsed from one of the option files somehow, so hopefully I don't need to wrangle any tex.
Dave
Quoted 5 lines
Quoted 5 lines
Signature · 2 lines
-- 2.1.0
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Quoted 8 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
This patchset provides a way of setting options on an incoming migration before the fd/process/socket has been created.
start qemu with -incoming defer
<use the monitor to set whatever you need>
migrate_incomingtheuri
Does this patch series break BDRV_O_INCOMING?
blockdev.c:blockdev_init():
if (runstate_check(RUN_STATE_INMIGRATE)) {
bdrv_flags |= BDRV_O_INCOMING;}
The block layer needs to know about incoming migrations to avoid accessing or caching disk images while the source host still has them open.
Stefan
Quoted 12 lines · * Stefan Hajnoczi ([email protected]) wrote:
Quoted 8 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
This patchset provides a way of setting options on an incoming migration before the fd/process/socket has been created.
start qemu with -incoming defer
<use the monitor to set whatever you need>
migrate_incomingtheuriDoes this patch series break
BDRV_O_INCOMING?
Quoted 8 lines
blockdev.c:blockdev_init():
if (runstate_check(RUN_STATE_INMIGRATE)) { bdrv_flags |= BDRV_O_INCOMING;}
The block layer needs to know about incoming migrations to avoid accessing or caching disk images while the source host still has them open.
No, it doesn't break it; it goes into RUN_STATE_INMIGRATE at startup in the same way that all existing -incoming's do; the state transitions should be identical.
Dave
Quoted 2 lines
Stefan
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Quoted 18 lines · Dr. David Alan Gilbert
Quoted 17 lines · * Stefan Hajnoczi ([email protected]) wrote:
Quoted 8 lines · Dr. David Alan Gilbert (git)
From: "Dr. David Alan Gilbert" <[email protected]>
This patchset provides a way of setting options on an incoming migration before the fd/process/socket has been created.
start qemu with -incoming defer
<use the monitor to set whatever you need>
migrate_incomingtheuriDoes this patch series break
BDRV_O_INCOMING?blockdev.c:blockdev_init():
if (runstate_check(RUN_STATE_INMIGRATE)) { bdrv_flags |= BDRV_O_INCOMING;}
Quoted 7 lines
Quoted 3 lines
The block layer needs to know about incoming migrations to avoid accessing or caching disk images while the source host still has them open.
No, it doesn't break it; it goes into
RUN_STATE_INMIGRATEat startup in the same way that all existing -incoming's do; the state transitions should be identical.
Great, thanks for explaining.
I have verified that vl.c:main() sets RUN_STATE_INMIGRATE in all cases:
case QEMU_OPTION_incoming:
incoming = optarg;
runstate_set(RUN_STATE_INMIGRATE);
break;Stefan