10:23 a.m.
These provide libcmpiutil support copying instances and raising indications, which are
required by the new MigrationIndication provider.
10:23 a.m.
New subject: [PATCH 1 of 2] [CU] Add dup_instance function to libcmpiutil
# HG changeset patch
# User Jay Gagnon <grendel(a)linux.vnet.ibm.com>
# Date 1201795147 18000
# Node ID ec7428b062f45db8eef730c5398d061d8731ce60
# Parent ab0e3b05a10e7b5db7eb493920f01bb402d2feae
[CU] Add dup_instance function to libcmpiutil
It appears there are some situations where we have to duplicate an instance; this seemed
like the kind of thing that would belong in libcmpiutil.
Signed-off-by: Jay Gagnon <grendel(a)linux.vnet.ibm.com>
diff -r ab0e3b05a10e -r ec7428b062f4 instance_util.c
--- a/instance_util.c Wed Jan 30 11:15:32 2008 -0800
+++ b/instance_util.c Thu Jan 31 10:59:07 2008 -0500
@@ -204,6 +204,28 @@ CMPIStatus cu_copy_prop(const CMPIBroker
return s;
}
+CMPIStatus cu_dup_instance(const CMPIBroker *broker,
+ CMPIInstance *src,
+ CMPIInstance **dest)
+{
+ CMPIStatus s = {CMPI_RC_OK, NULL};
+ CMPIObjectPath *ref;
+
+ ref = CMGetObjectPath(src, NULL);
+ if ((s.rc != CMPI_RC_OK) || CMIsNullObject(ref)) {
+ cu_statusf(broker, &s,
+ CMPI_RC_ERR_FAILED,
+ "Could not get objectpath from instance");
+ goto out;
+ }
+
+ *dest = NULL;
+ *dest = CMNewInstance(broker, ref, &s);
+
+ out:
+ return s;
+}
+
/*
* Local Variables:
* mode: C
diff -r ab0e3b05a10e -r ec7428b062f4 libcmpiutil.h
--- a/libcmpiutil.h Wed Jan 30 11:15:32 2008 -0800
+++ b/libcmpiutil.h Thu Jan 31 10:59:07 2008 -0500
@@ -164,6 +164,18 @@ CMPIrc cu_get_u16_path(const CMPIObjectP
CMPIrc cu_get_u16_path(const CMPIObjectPath *reference,
const char *key,
uint16_t *target);
+
+/**
+ * Create a copy of an instance
+ *
+ * @param src Source instance
+ * @param dest Destination instance
+ * @returns {CMPI_RC_OK, NULL} if success, CMPI_RC ERR_FAILED and
+ * error message otherwise
+ */
+CMPIStatus cu_dup_instance(const CMPIBroker *broker,
+ CMPIInstance *src,
+ CMPIInstance **dest);
/* Forward declaration */
struct inst_list;
10:59 a.m.
New subject: [PATCH 1 of 2] [CU] Add dup_instance function to libcmpiutil
JG> +CMPIStatus cu_dup_instance(const CMPIBroker *broker,
JG> + CMPIInstance *src,
JG> + CMPIInstance **dest)
JG> +{
JG> + CMPIStatus s = {CMPI_RC_OK, NULL};
JG> + CMPIObjectPath *ref;
JG> +
JG> + ref = CMGetObjectPath(src, NULL);
JG> + if ((s.rc != CMPI_RC_OK) || CMIsNullObject(ref)) {
JG> + cu_statusf(broker, &s,
JG> + CMPI_RC_ERR_FAILED,
JG> + "Could not get objectpath from instance");
JG> + goto out;
JG> + }
JG> +
JG> + *dest = NULL;
Why set it to NULL and then set it to the instance right after this?
JG> + *dest = CMNewInstance(broker, ref, &s);
JG> +
JG> + out:
JG> + return s;
JG> +}
What about all the properties of the instance? This gets a new
instance structure, yes, but all it has is an object path. Maybe that
is all you need at the moment, but I don't think this does what other
consumers of the library would think by the name :)
It should be easy to loop through all the properties and copy them
over. It will look similar to the array copy/filter code in Kaitlin's
libcmpiutil CO patch.
--
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms(a)us.ibm.com
11:11 a.m.
New subject: [PATCH 1 of 2] [CU] Add dup_instance function to libcmpiutil
Dan Smith wrote:
JG> +CMPIStatus cu_dup_instance(const CMPIBroker *broker,
JG> + CMPIInstance *src,
JG> + CMPIInstance **dest)
JG> +{
JG> + CMPIStatus s = {CMPI_RC_OK, NULL};
JG> + CMPIObjectPath *ref;
JG> +
JG> + ref = CMGetObjectPath(src, NULL);
JG> + if ((s.rc != CMPI_RC_OK) || CMIsNullObject(ref)) {
JG> + cu_statusf(broker, &s,
JG> + CMPI_RC_ERR_FAILED,
JG> + "Could not get objectpath from instance");
JG> + goto out;
JG> + }
JG> +
JG> + *dest = NULL;
Why set it to NULL and then set it to the instance right after this?
I think I was going with "not sure what *dest becomes if NewInstance"
fails, and so I figured better safe than sorry. I'd rather have it NULL
than undefined.
JG> + *dest = CMNewInstance(broker, ref, &s);
JG> +
JG> + out:
JG> + return s;
JG> +}
What about all the properties of the instance? This gets a new
instance structure, yes, but all it has is an object path. Maybe that
is all you need at the moment, but I don't think this does what other
consumers of the library would think by the name :)
It should be easy to loop through all the properties and copy them
over. It will look similar to the array copy/filter code in Kaitlin's
libcmpiutil CO patch.
I'm pleading CIM ignorance on this one. I thought the ObjectPath
encompassed the properties, and thought that my testing showed it
happening (I don't manually move the JobState over but it is est), but I
could have been wrong.
--
-Jay
2:13 p.m.
New subject: [PATCH 1 of 2] [CU] Add dup_instance function to libcmpiutil
JG> I'm pleading CIM ignorance on this one. I thought the ObjectPath
JG> encompassed the properties, and thought that my testing showed it
JG> happening (I don't manually move the JobState over but it is est),
JG> but I could have been wrong.
You're only half ignorant here :)
The ObjectPath has the key properties, but not the rest.
--
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms(a)us.ibm.com
4:58 a.m.
New subject: [PATCH 1 of 2] [CU] Add dup_instance function to libcmpiutil
Dan Smith wrote:
JG> I'm pleading CIM ignorance on this one. I thought the
ObjectPath
JG> encompassed the properties, and thought that my testing showed it
JG> happening (I don't manually move the JobState over but it is est),
JG> but I could have been wrong.
You're only half ignorant here :)
The ObjectPath has the key properties, but not the rest.
But it can also be, that some of the none-key properties contain a
value. This happens, if within the DMTF's mof files a default value was
defined. Might this be for JobState ?
--
Regards
Heidi Eckhart
Software Engineer
IBM Linux Technology Center - Open Hypervisor
8:21 a.m.
New subject: [PATCH 1 of 2] [CU] Add dup_instance function to libcmpiutil
Heidi Eckhart wrote:
Dan Smith wrote:
> JG> I'm pleading CIM ignorance on this one. I thought the ObjectPath
> JG> encompassed the properties, and thought that my testing showed it
> JG> happening (I don't manually move the JobState over but it is est),
> JG> but I could have been wrong.
>
> You're only half ignorant here :)
>
> The ObjectPath has the key properties, but not the rest.
>
>
But it can also be, that some of the none-key properties contain a
value. This happens, if within the DMTF's mof files a default value
was defined. Might this be for JobState ?
Right on the money. I was looking for something that didn't appear to
copy correctly, in order to confirm the fix when I'm done working on
cu_dup_instance, and JobState was the first one I found.
--
-Jay
10:23 a.m.
New subject: [PATCH 2 of 2] [CU] Improve std_indication's raise functionality
# HG changeset patch
# User Jay Gagnon <grendel(a)linux.vnet.ibm.com>
# Date 1201882988 18000
# Node ID 8669586635358e4858e56fe2dec8587af784cd47
# Parent ec7428b062f45db8eef730c5398d061d8731ce60
[CU] Improve std_indication's raise functionality
Turns out STDI_IndicationMIStub was not creating the CMPIFooMIFTs correctly; this patch
fixes that. It also adds a default_raise function to std_indication. This is for
indication providers that don't need to do anything other than call
CBDeliverIndication, which will most likely be the majority.
Signed-off-by: Jay Gagnon <grendel(a)linux.vnet.ibm.com>
diff -r ec7428b062f4 -r 866958663535 std_indication.c
--- a/std_indication.c Thu Jan 31 10:59:07 2008 -0500
+++ b/std_indication.c Fri Feb 01 11:23:08 2008 -0500
@@ -19,6 +19,7 @@
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#include <string.h>
+#include <stdbool.h>
#include <cmpidt.h>
#include <cmpift.h>
@@ -40,19 +41,40 @@ static CMPIStatus trigger(struct std_ind
return ctx->handler->trigger_fn(context);
}
+static CMPIStatus default_raise(const CMPIBroker *broker,
+ const CMPIContext *context,
+ CMPIInstance *ind)
+{
+ CMPIObjectPath *ref;
+ CMPIStatus s = {CMPI_RC_OK, NULL};
+
+ ref = CMGetObjectPath(ind, NULL);
+
+ CBDeliverIndication(broker,
+ context,
+ NAMESPACE(ref),
+ ind);
+ return s;
+}
+
static CMPIStatus raise(struct std_indication_ctx *ctx,
const CMPIContext *context,
const CMPIArgs *argsin)
{
CMPIInstance *inst;
- if (ctx->handler->raise_fn == NULL)
- return (CMPIStatus){CMPI_RC_OK, NULL};
+ if (!ctx->enabled) {
+ CU_DEBUG("Indication disabled, not raising.");
+ return (CMPIStatus) {CMPI_RC_OK, NULL};
+ }
if (cu_get_inst_arg(argsin, "Indication", &inst) != CMPI_RC_OK)
return (CMPIStatus){CMPI_RC_ERR_FAILED, NULL};
- return ctx->handler->raise_fn(context, inst);
+ if (ctx->handler->raise_fn == NULL)
+ return default_raise(ctx->brkr, context, inst);
+
+ return ctx->handler->raise_fn(ctx->brkr, context, inst);
}
CMPIStatus stdi_handler(CMPIMethodMI *self,
diff -r ec7428b062f4 -r 866958663535 std_indication.h
--- a/std_indication.h Thu Jan 31 10:59:07 2008 -0500
+++ b/std_indication.h Fri Feb 01 11:23:08 2008 -0500
@@ -24,6 +24,7 @@
#include <cmpidt.h>
#include <cmpift.h>
#include <cmpimacs.h>
+#include <stdio.h>
#include "libcmpiutil.h"
#include "std_invokemethod.h"
@@ -51,7 +52,8 @@ CMPIStatus stdi_cleanup(CMPIMethodMI *se
const CMPIContext *context,
CMPIBoolean terminating);
-typedef CMPIStatus (*raise_indication_t)(const CMPIContext *ctx,
+typedef CMPIStatus (*raise_indication_t)(const CMPIBroker *broker,
+ const CMPIContext *ctx,
const CMPIInstance *ind);
typedef CMPIStatus (*trigger_indication_t)(const CMPIContext *ctx);
@@ -64,14 +66,44 @@ struct std_indication_ctx {
struct std_indication_ctx {
const CMPIBroker *brkr;
struct std_indication_handler *handler;
+ bool enabled;
};
#define STDI_IndicationMIStub(pfx, pn, _broker, hook, _handler) \
+ static struct std_indication_ctx _ctx = { \
+ .brkr = NULL, \
+ .handler = _handler, \
+ .enabled = false, \
+ }; \
+ \
+ static CMPIIndicationMIFT indMIFT__ = { \
+ CMPICurrentVersion, \
+ CMPICurrentVersion, \
+ "Indication" #pn, \
+ pfx##IndicationCleanup, \
+ pfx##AuthorizeFilter, \
+ pfx##MustPoll, \
+ pfx##ActivateFilter, \
+ pfx##DeActivateFilter, \
+ CMIndicationMIStubExtensions(pfx) \
+ }; \
CMPIIndicationMI * \
pn##_Create_IndicationMI(const CMPIBroker *, \
- const CMPIContext *, \
- CMPIStatus *); \
- CMIndicationMIStub(pfx, pn, _broker, hook); \
+ const CMPIContext *, \
+ CMPIStatus *); \
+ CMPIIndicationMI * \
+ pn##_Create_IndicationMI(const CMPIBroker *brkr, \
+ const CMPIContext *ctx, \
+ CMPIStatus *rc) { \
+ static CMPIIndicationMI mi = { \
+ &_ctx, \
+ &indMIFT__, \
+ }; \
+ _ctx.brkr = brkr; \
+ _broker = brkr; \
+ hook; \
+ return &mi; \
+ } \
\
static CMPIMethodMIFT methMIFT__ = { \
CMPICurrentVersion, \
@@ -88,13 +120,11 @@ struct std_indication_ctx {
CMPIMethodMI *pn##_Create_MethodMI(const CMPIBroker *brkr, \
const CMPIContext *ctx, \
CMPIStatus *rc) { \
- static struct std_indication_ctx _ctx; \
static CMPIMethodMI mi = { \
&_ctx, \
&methMIFT__, \
}; \
_ctx.brkr = brkr; \
- _ctx.handler = _handler; \
_broker = brkr; \
hook; \
return &mi; \
11:14 a.m.
New subject: [PATCH 2 of 2] [CU] Improve std_indication's raise functionality
JG> # HG changeset patch
JG> # User Jay Gagnon <grendel(a)linux.vnet.ibm.com>
JG> # Date 1201882988 18000
JG> # Node ID 8669586635358e4858e56fe2dec8587af784cd47
JG> # Parent ec7428b062f45db8eef730c5398d061d8731ce60
JG> [CU] Improve std_indication's raise functionality
This one looks okay to me, and I think I can apply it independently of
the other, correct?
--
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms(a)us.ibm.com
9:02 a.m.
New subject: [PATCH 2 of 2] [CU] Improve std_indication's raise functionality
Dan Smith wrote:
JG> # HG changeset patch
JG> # User Jay Gagnon <grendel(a)linux.vnet.ibm.com>
JG> # Date 1201882988 18000
JG> # Node ID 8669586635358e4858e56fe2dec8587af784cd47
JG> # Parent ec7428b062f45db8eef730c5398d061d8731ce60
JG> [CU] Improve std_indication's raise functionality
This one looks okay to me, and I think I can apply it independently of
the other, correct?
------------------------------------------------------------------------
_______________________________________________
Libvirt-cim mailing list
Libvirt-cim(a)redhat.com
https://www.redhat.com/mailman/listinfo/libvirt-cim
Yea, this should be able to go
in independently. Migration is the only
indication that uses raise, so we won't be pulling the rug out from
under anybody.
--
-Jay
11:06 a.m.
This set looks good. +1
--
Kaitlin Rupert
IBM Linux Technology Center
kaitlin(a)linux.vnet.ibm.com
6118
days inactive
6122
days old
10 comments
4 participants
participants (4)
-
Dan Smith -
Heidi Eckhart -
Jay Gagnon -
Kaitlin Rupert