[libvirt] [PATCHv2 0/4] qemu: fix broken RTC_CHANGE event when clock offset='variable'
by Laine Stump
(the cover letter from V1 has been preserved here with small updates)
This patch series is all about:
https://bugzilla.redhat.com/show_bug.cgi?id=964177
which points out a flaw in qemu's RTC_CHANGE event when the base for
the guest's real time clock was given as an explicit date (rather than
as 'utc' or 'localtime'). Patch 3/4 explains what qemu does, and how
this patch fixes it (for the case of basis='utc' - an additional fix
to properly support basis='localtime', both for RTC_CHANGE and when
migrating across timezone/DST boundaries, is in Patch 4/4). (NB: the
flawed RTC_CHANGE offset has been in QEMU for so long now that it has
been documented and cannot be changed, so libvirt must work around it)
In the meantime, the fix for basis='localtime' required that we learn
the offset of localtime from utc, and that seems like something we
might want to do for other reasons, so Patch 1/4 adds a new utility
function do get that value that is (I hope!) POSIX compliant.
Since the original patch to fix this problem was incorrect, and the
new patch doesn't use any of its code, Patch 2/4 reverts that patch
completely.
Note that I have tested this patch by starting a domain with several
variations of <clock> parameters (in a locale that is currently at
UTC+3) and using the following short script to add and remove seconds
from the guest's RTC while watching both the <clock> line in
/var/run/libvirt/qemu/$domain.xml and the offset value sent in
libvirt's VIR_DOMAIN_EVENT_ID_RTC_CHANGE event (using
examples/domain-events/events-c/event-test from a libvirt source tree
that has been built). All cases appear to maintain the adjustment in
the status properly, as well as sending the correct value to any event
handler.
Here is the script I used to add/remove time from the RTC:
#!/bin/sh
old=$(hwclock --show | cut -f1-7 -d' ')
oldabs=$(date +%s --date="$old")
newabs=$(expr $oldabs + $1)
new=$(date --date="@$newabs")
echo Old: $oldabs $old
echo New: $newabs $new
hwclock --set --date="@$newabs"
Laine Stump (4):
util: new function virTimeLocalOffsetFromUTC
Revert "qemu: Report the offset from host UTC for RTC_CHANGE event"
qemu: fix RTC_CHANGE event for <clock offset='variable' basis='utc'/>
qemu: fix <clock offset='variable' basis='localtime'/>
src/conf/domain_conf.c | 32 +++++++++++++-------------------
src/conf/domain_conf.h | 8 +++++---
src/libvirt_private.syms | 1 +
src/qemu/qemu_command.c | 43 ++++++++++++++++++++++++++++++-------------
src/qemu/qemu_process.c | 34 +++++++++++++++++++---------------
src/util/virtime.c | 41 ++++++++++++++++++++++++++++++++++++++++-
src/util/virtime.h | 5 +++--
tests/virtimetest.c | 44 ++++++++++++++++++++++++++++++++++++++++++++
8 files changed, 155 insertions(+), 53 deletions(-)
--
1.9.0
10 years, 1 month
[libvirt] [PATCH] build: avoid compiler warning on 32-bit platform
by Eric Blake
On a 32-bit platform:
virstringtest.c: In function 'mymain':
virstringtest.c:673: warning: this decimal constant is unsigned only in ISO C90
I already had a comment in the file about the 64-bit counterpart;
the easiest fix was to make both sites use the standardized macro
that is guaranteed to work.
* tests/virstringtest.c (mymain): Minimum signed integers are a pain.
Signed-off-by: Eric Blake <eblake(a)redhat.com>
---
Pushing under the build-breaker rule.
tests/virstringtest.c | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/tests/virstringtest.c b/tests/virstringtest.c
index 1e330f9..96c3784 100644
--- a/tests/virstringtest.c
+++ b/tests/virstringtest.c
@@ -670,7 +670,7 @@ mymain(void)
-1LL, 0, 18446744073709551615ULL, 0);
TEST_STRTOL("-2147483647", NULL, -2147483647, 0, 2147483649U, 0,
-2147483647LL, 0, 18446744071562067969ULL, 0);
- TEST_STRTOL("-2147483648", NULL, -2147483648, 0, 2147483648U, 0,
+ TEST_STRTOL("-2147483648", NULL, INT32_MIN, 0, 2147483648U, 0,
-2147483648LL, 0, 18446744071562067968ULL, 0);
TEST_STRTOL("-2147483649", NULL, 0, -1, 2147483647U, 0,
-2147483649LL, 0, 18446744071562067967ULL, 0);
@@ -680,10 +680,8 @@ mymain(void)
-4294967296LL, 0, 18446744069414584320ULL, 0);
TEST_STRTOL("-9223372036854775807", NULL, 0, -1, 0U, -1,
-9223372036854775807LL, 0, 9223372036854775809ULL, 0);
- /* Bah, stupid gcc warning about -9223372036854775808LL being an
- * unrepresentable integer constant */
TEST_STRTOL("-9223372036854775808", NULL, 0, -1, 0U, -1,
- 0x8000000000000000LL, 0, 9223372036854775808ULL, 0);
+ INT64_MIN, 0, 9223372036854775808ULL, 0);
TEST_STRTOL("-9223372036854775809", NULL, 0, -1, 0U, -1,
0LL, -1, 9223372036854775807ULL, 0);
TEST_STRTOL("-18446744073709551615", NULL, 0, -1, 0U, -1,
--
1.9.3
10 years, 2 months
[libvirt] [PATCH] libxl: add support for cache=directsync to to libxl_device_disk
by Olaf Hering
Up to now xend and libxl did not have any way and need to specify a cache=
mode for block devices. At least the libxl driver did just ignore the
cachemode value.
With xen-4.5 a new knob was added to libxl which instructs qemu to use
threads=native cache=directsync in the qdisk driver. This mode was disabled
early due to a pvops kernel bug. As a workaround cache=writeback was set as
default. The kernel bug is long fixed. I assume the forward ported xenlinux
kernel never had this bug.
This change exposes the knob to libvirt. If cache=directsync is set qemu is
instructed to use native AIO.
This change also lays the ground work to set cache=unsafe at some point.
It is currently under investigation whether libxl should expose such knob.
Signed-off-by: Olaf Hering <olaf(a)aepfle.de>
---
src/libxl/libxl_conf.c | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c
index 4cb062e..79f4a82 100644
--- a/src/libxl/libxl_conf.c
+++ b/src/libxl/libxl_conf.c
@@ -739,6 +739,20 @@ libxlDiskSetDiscard(libxl_device_disk *x_disk, int discard)
#endif
}
+static void
+libxlDiskSetCacheMode(libxl_device_disk *x_disk, int cachemode)
+{
+ switch (cachemode) {
+#if defined(LIBXL_HAVE_DEVICE_DISK_DIRECT_IO_SAFE)
+ case VIR_DOMAIN_DISK_CACHE_DIRECTSYNC:
+ x_disk->direct_io_safe = 1;
+ break;
+#endif
+ default:
+ break;
+ }
+}
+
int
libxlMakeDisk(virDomainDiskDefPtr l_disk, libxl_device_disk *x_disk)
@@ -855,6 +869,7 @@ libxlMakeDisk(virDomainDiskDefPtr l_disk, libxl_device_disk *x_disk)
x_disk->readwrite = !l_disk->readonly;
x_disk->is_cdrom = l_disk->device == VIR_DOMAIN_DISK_DEVICE_CDROM ? 1 : 0;
libxlDiskSetDiscard(x_disk, l_disk->discard);
+ libxlDiskSetCacheMode(x_disk, l_disk->cachemode);
/* An empty CDROM must have the empty format, otherwise libxl fails. */
if (x_disk->is_cdrom && !x_disk->pdev_path)
x_disk->format = LIBXL_DISK_FORMAT_EMPTY;
10 years, 2 months
[libvirt] [libvirt-python PATCH] fix leak in memoryStats with older python
by Martin Kletzander
libvirt_virDomainMemoryStats() function creates a dictionary without
any checks whether the additions were successful, whether the python
objects were created and, most importantly, without decrementing the
reference count on the objects added to the dictionary. This is
somehow not an issue with current upstream versions, however with
python 2.6 this exposes a leak in our bindings. The following patch
works on both old and new CPython versions and is already used in
other parts of the code, so it's also most straightforward.
Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com>
---
libvirt-override.c | 68 +++++++++++++++++++++++++++++++++++-------------------
1 file changed, 44 insertions(+), 24 deletions(-)
diff --git a/libvirt-override.c b/libvirt-override.c
index a7a6213..8fd856b 100644
--- a/libvirt-override.c
+++ b/libvirt-override.c
@@ -734,6 +734,7 @@ libvirt_virDomainMemoryStats(PyObject *self ATTRIBUTE_UNUSED, PyObject *args) {
size_t i;
virDomainMemoryStatStruct stats[VIR_DOMAIN_MEMORY_STAT_NR];
PyObject *info;
+ PyObject *key = NULL, *val = NULL;
if (!PyArg_ParseTuple(args, (char *)"O:virDomainMemoryStats", &pyobj_domain))
return NULL;
@@ -749,31 +750,50 @@ libvirt_virDomainMemoryStats(PyObject *self ATTRIBUTE_UNUSED, PyObject *args) {
return VIR_PY_NONE;
for (i = 0; i < nr_stats; i++) {
- if (stats[i].tag == VIR_DOMAIN_MEMORY_STAT_SWAP_IN)
- PyDict_SetItem(info, libvirt_constcharPtrWrap("swap_in"),
- libvirt_ulonglongWrap(stats[i].val));
- else if (stats[i].tag == VIR_DOMAIN_MEMORY_STAT_SWAP_OUT)
- PyDict_SetItem(info, libvirt_constcharPtrWrap("swap_out"),
- libvirt_ulonglongWrap(stats[i].val));
- else if (stats[i].tag == VIR_DOMAIN_MEMORY_STAT_MAJOR_FAULT)
- PyDict_SetItem(info, libvirt_constcharPtrWrap("major_fault"),
- libvirt_ulonglongWrap(stats[i].val));
- else if (stats[i].tag == VIR_DOMAIN_MEMORY_STAT_MINOR_FAULT)
- PyDict_SetItem(info, libvirt_constcharPtrWrap("minor_fault"),
- libvirt_ulonglongWrap(stats[i].val));
- else if (stats[i].tag == VIR_DOMAIN_MEMORY_STAT_UNUSED)
- PyDict_SetItem(info, libvirt_constcharPtrWrap("unused"),
- libvirt_ulonglongWrap(stats[i].val));
- else if (stats[i].tag == VIR_DOMAIN_MEMORY_STAT_AVAILABLE)
- PyDict_SetItem(info, libvirt_constcharPtrWrap("available"),
- libvirt_ulonglongWrap(stats[i].val));
- else if (stats[i].tag == VIR_DOMAIN_MEMORY_STAT_ACTUAL_BALLOON)
- PyDict_SetItem(info, libvirt_constcharPtrWrap("actual"),
- libvirt_ulonglongWrap(stats[i].val));
- else if (stats[i].tag == VIR_DOMAIN_MEMORY_STAT_RSS)
- PyDict_SetItem(info, libvirt_constcharPtrWrap("rss"),
- libvirt_ulonglongWrap(stats[i].val));
+ switch (stats[i].tag) {
+ case VIR_DOMAIN_MEMORY_STAT_SWAP_IN:
+ key = libvirt_constcharPtrWrap("swap_in");
+ break;
+ case VIR_DOMAIN_MEMORY_STAT_SWAP_OUT:
+ key = libvirt_constcharPtrWrap("swap_out");
+ break;
+ case VIR_DOMAIN_MEMORY_STAT_MAJOR_FAULT:
+ key = libvirt_constcharPtrWrap("major_fault");
+ break;
+ case VIR_DOMAIN_MEMORY_STAT_MINOR_FAULT:
+ key = libvirt_constcharPtrWrap("minor_fault");
+ break;
+ case VIR_DOMAIN_MEMORY_STAT_UNUSED:
+ key = libvirt_constcharPtrWrap("unused");
+ break;
+ case VIR_DOMAIN_MEMORY_STAT_AVAILABLE:
+ key = libvirt_constcharPtrWrap("available");
+ break;
+ case VIR_DOMAIN_MEMORY_STAT_ACTUAL_BALLOON:
+ key = libvirt_constcharPtrWrap("actual");
+ break;
+ case VIR_DOMAIN_MEMORY_STAT_RSS:
+ key = libvirt_constcharPtrWrap("rss");
+ break;
+ default:
+ continue;
+ }
+ val = libvirt_ulonglongWrap(stats[i].val);
+
+ if (!key || !val || PyDict_SetItem(info, key, val) < 0) {
+ Py_DECREF(info);
+ info = NULL;
+ goto cleanup;
+ }
+ Py_DECREF(key);
+ Py_DECREF(val);
+ key = NULL;
+ val = NULL;
}
+
+ cleanup:
+ Py_XDECREF(key);
+ Py_XDECREF(val);
return info;
}
--
1.9.3
10 years, 2 months
[libvirt] Host OS, Storage Info
by Sijo Jose
Hi,
Is there any way to get host OS information and host Storage in formations
using libvirt API...?
Rgds
-Sijo
10 years, 2 months
[libvirt] [libvirt-glib] [PATCH] GVirDomainSnapshot: add gvir_domain_snapshot_delete
by Timm Bäder
---
libvirt-gobject/libvirt-gobject-domain-snapshot.c | 21 +++++++++++++++++++++
libvirt-gobject/libvirt-gobject-domain-snapshot.h | 3 +++
libvirt-gobject/libvirt-gobject.sym | 5 +++++
3 files changed, 29 insertions(+)
diff --git a/libvirt-gobject/libvirt-gobject-domain-snapshot.c b/libvirt-gobject/libvirt-gobject-domain-snapshot.c
index ab23342..f46c99b 100644
--- a/libvirt-gobject/libvirt-gobject-domain-snapshot.c
+++ b/libvirt-gobject/libvirt-gobject-domain-snapshot.c
@@ -206,3 +206,24 @@ GVirConfigDomainSnapshot *gvir_domain_snapshot_get_config
free(xml);
return conf;
}
+
+/**
+ * gvir_domain_snapshot_delete:
+ * @snapshot: the domain_snapshot
+ * @error: (allow-none): Place-holder for error or NULL
+ */
+void gvir_domain_snapshot_delete(GVirDomainSnapshot *snapshot, GError **error)
+{
+ GVirDomainSnapshotPrivate *priv;
+ int status;
+
+ g_return_if_fail(GVIR_IS_DOMAIN_SNAPSHOT (snapshot));
+ g_return_if_fail(error == NULL || *error == NULL);
+
+ priv = snapshot->priv;
+ status = virDomainSnapshotDelete(priv->handle, 0);
+ if (status < 0) {
+ gvir_set_error_literal(error, GVIR_DOMAIN_SNAPSHOT_ERROR, 0,
+ "Unable to delete snapshot");
+ }
+}
diff --git a/libvirt-gobject/libvirt-gobject-domain-snapshot.h b/libvirt-gobject/libvirt-gobject-domain-snapshot.h
index 5bd827c..575f56f 100644
--- a/libvirt-gobject/libvirt-gobject-domain-snapshot.h
+++ b/libvirt-gobject/libvirt-gobject-domain-snapshot.h
@@ -69,6 +69,9 @@ GVirConfigDomainSnapshot *gvir_domain_snapshot_get_config
guint flags,
GError **err);
+void gvir_domain_snapshot_delete (GVirDomainSnapshot *snapshot, GError **error);
+
+
G_END_DECLS
#endif /* __LIBVIRT_GOBJECT_DOMAIN_SNAPSHOT_H__ */
diff --git a/libvirt-gobject/libvirt-gobject.sym b/libvirt-gobject/libvirt-gobject.sym
index f2419ac..5d12a6a 100644
--- a/libvirt-gobject/libvirt-gobject.sym
+++ b/libvirt-gobject/libvirt-gobject.sym
@@ -234,4 +234,9 @@ LIBVIRT_GOBJECT_0.1.5 {
gvir_connection_open_read_only_finish;
} LIBVIRT_GOBJECT_0.1.4;
+LIBVIRT_GOBJECT_0.1.9 {
+ global:
+ gvir_domain_shapshot_delete;
+} LIBVIRT_GOBJECT_0.1.5;
+
# .... define new API here using predicted next version number ....
--
1.9.3
10 years, 2 months
[libvirt] GSoC student introduction and work plan
by Taowei Luo
Hi, everyone, My name is Taowei Luo. I'm one of the students working for
GSoC this summer. My project is rewriting the vbox driver.
Before coding, I really need some discussions about how to implement it.
Here I post my first version of plan. I had show as much details as I
thought about in this plan. If anyone have ideas or suggestions about my
work, I would like to hear. My nick name in #virt IRC channel is uaedante
(or dante I used before). If I am offline in IRC, please contract me with
my email address uaedante(a)gmail.com.
Regards,
Taowei.
Plan to rewrite vbox driver
Introduction:
The existing vbox driver codes use vbox API directly. But the vbox API is
incompatible between different versions. To support all vbox API versions,
the vbox_tmpl.c file provides a template to implement all vbox drivers.
Each specified version of vbox driver includes the vbox_tmpl.c file, with
the definition in the corresponding head file, they work well. As the vbox
API defined in head files is conflict between different versions, even some
behavior of vbox APIs are changed. The vbox_tmpl.c contains lots of codes
such as “#if VBOX_API_VERSION < 4001000”. These codes is hard to manage or
adding new features.
Plan:
The core idea of my rewriting work is to add a middle layer to uniform the
API changes. This modification will not change any functions’ behavior or
sematic.
The middle layer has a global structure which contains API changes. Like
this:
struct vboxUniformedAPI {
A (*funcA)(A1 a, ….);
B (*funcB)(B1 a, ….);
}
If there is a code segment:
#if VBOX_API_VERSION < 4001000
adapter->vtbl->GetHostInterface(adapter, &hostIntUtf16);
#else /* VBOX_API_VERSION >= 4001000 */
adapter->vtbl->GetBridgedInterface(adapter, &hostIntUtf16);
#endif /* VBOX_API_VERSION >= 4001000 */
This will result to a variable in vboxUniformedAPI structure:
RetType (*GetInterface)(INetworkAdapter *a, PRUnichar *b)
When using the vbox under 4.1, this variable will points to a function that
calls GetHostInterface. When using the vbox later than 4.1 and the variable
will points to a function which calls GetBridgedInterface.
When replace these code segments with a single function, we need to do
living variable analyze and find out what is used and what is changed in
these codes. I plan to do these analyze manually.
All conflict in function prototype or function implementation will be fixed
up though this method.
When meeting a conflict in variable type, like:
#if VBOX_API_VERSION < 4001000
TypeA var;
#else /* VBOX_API_VERSION >= 4001000 */
TypeB var;
#endif /* VBOX_API_VERSION >= 4001000 */
I will add a union like this:
Union TypeC{
TypeA a;// used in some versions
TypeB b;// used in some other versions
}
If there is any operation using variables with such a union type (or
functions use it as an argument or return value), I will add a new function
in vboxUniformedAPI to handle this operation. These functions use TypeC as
parameter type, dispatch it depends on the vbox version and return TypeC
value to conceal type conflict.
We still have another kind of conflict. There are some versions have more
variables, operations or function calls than other versions. I will reserve
all variables even though they may not be used in some situation. And put
an empty function if they have nothing to do in the specified function
step. (I have a backup solution here. That is adding some flags to indicate
whether the procession is necessary in the API version and prevent upper
layer from calling a nonexistent function.)
Expected result:
After rewriting with the rules above, we will have a new vbox_tmpl.c. It
gathers version specified code and common code respectively. The common
codes use stable vbox API or functions in vboxUniformedAPI. A function
named InstallvboxUnifromedAPI will fill the vboxUniformedAPI structure
depends on the current vbox API version. In final step, I will put all vbox
APIs into the vboxUniformedAPI and make the vbox driver codes use my API
only. If this is done, the vbox driver will not be complied for each
version anymore.
Implementation:
First, define the vboxUniformedAPI structure in vbox_tmpl.c, and write the
InstallvboxUnifromedAPI function for every vbox API version.
Next, rewrite the functions which have conflict codes one by one. Each time
rewrite a new function, we can do some tests (here, I am not sure what kind
of tests is enough). When rewriting one function in vbox_tmpl.c, the others
will not change.
At this point, there will split version specified codes and common codes
But we still need to build the whole vbox_tmpl.c for each vbox API version.
If we want a single object file be generated by vbox_tmpl.c, we need to
prevent vbox_tmpl.c from calling vbox API directly. This could be done by
putting all of the vbox API calls into the middle vboxUniformedAPI layer,
and moving common codes into vbox_common.c. The vbox_common.c should only
use APIs in vboxUniformedAPI and will be complied only once.
10 years, 2 months
[libvirt] libvirt: Xen Driver error : Domain not found: xenGetDomainDefForName
by deepanshus
Hi,
I have successfully installed xen 3.11 kernel but when I am running the vm I am
getting this error:
libvirt: Xen Driver error : Domain not found: xenGetDomainDefForName
I am sending the xend log file as well error file as an attachment.
Thanks And Regards,
Deepanshu Saxena
-------------------------------------------------------------------------------------------------------------------------------
[ C-DAC is on Social-Media too. Kindly follow us at:
Facebook: https://www.facebook.com/CDACINDIA & Twitter: @cdacindia ]
This e-mail is for the sole use of the intended recipient(s) and may
contain confidential and privileged information. If you are not the
intended recipient, please contact the sender by reply e-mail and destroy
all copies and the original message. Any unauthorized review, use,
disclosure, dissemination, forwarding, printing or copying of this email
is strictly prohibited and appropriate legal action will be taken.
-------------------------------------------------------------------------------------------------------------------------------
10 years, 2 months
