1:14 a.m.
This patch adds the lxc_driver source files.
--
Best Regards,
Dave Leskovec
IBM Linux Technology Center
Open Virtualization
11:11 a.m.
On Wed, Mar 19, 2008 at 11:14:59PM -0700, Dave Leskovec wrote:
This patch adds the lxc_driver source files.
[...]
+static int lxcCheckContainerSupport( void )
+{
+ int rc = 0;
+ int flags = CLONE_NEWPID|CLONE_NEWNS|CLONE_NEWUTS|CLONE_NEWUSER|
+ CLONE_NEWIPC|SIGCHLD;
+ int cpid;
+ char *childStack;
+ char *stack;
+ int childStatus;
+
+ stack = malloc(getpagesize() * 4);
+ if(!stack) {
+ DEBUG0("Unable to allocate stack");
+ rc = -1;
+ goto check_complete;
+ }
+
+ childStack = stack + (getpagesize() * 4);
+
+ cpid = clone(lxcDummyChild, childStack, flags, NULL);
+ if ((0 > cpid) && (EINVAL == errno)) {
+ DEBUG0("clone call returned EINVAL, container support is not
enabled");
+ rc = -1;
haha, I would have expected a checking of /proc or something similar.
That test could still fail, say if the kernel started to disagree on
stack of only 4 pages for example.
+ } else {
+ waitpid(cpid, &childStatus, 0);
+ }
+
+ free(stack);
+
+check_complete:
+ return rc;
+}
[...]
+static virDrvOpenStatus lxcOpen(virConnectPtr conn,
+ xmlURIPtr uri,
+ virConnectAuthPtr auth ATTRIBUTE_UNUSED,
+ int flags ATTRIBUTE_UNUSED)
+{
+ uid_t uid = getuid();
+
+ /* Check that the user is root */
+ if (0 != uid) {
+ goto declineConnection;
+ }
so it's restricted to root, it's probably fine, as we can go though the
daemon for normal users, ssuming they get authenticated.
[...]
+static int lxcListDomains(virConnectPtr conn, int *ids, int nids)
+{
+ lxc_driver_t *driver = (lxc_driver_t *)conn->privateData;
+ lxc_vm_t *vm;
+ int numDoms = 0;
+
+ for (vm = driver->vms; vm && (numDoms < nids); vm = vm->next) {
+ if (lxcIsActiveVM(vm)) {
+ ids[numDoms] = vm->def->id;
+ numDoms++;
+ }
+ }
+
+ return numDoms;
+}
so we can only list domains created by this libvirt instance, right ?
Or I'm missing something, I assume virsh list works but I don't see how.
Except this bit I don't understand this looks fine to me
Index: b/src/lxc_driver.h
Looks fine,
thanks !
Daniel
--
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard | virtualization library http://libvirt.org/
veillard(a)redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
11:14 a.m.
On Thu, Mar 20, 2008 at 12:11:39PM -0400, Daniel Veillard wrote:
On Wed, Mar 19, 2008 at 11:14:59PM -0700, Dave Leskovec wrote:
> This patch adds the lxc_driver source files.
[...]
> +static int lxcCheckContainerSupport( void )
> +{
> + int rc = 0;
> + int flags = CLONE_NEWPID|CLONE_NEWNS|CLONE_NEWUTS|CLONE_NEWUSER|
> + CLONE_NEWIPC|SIGCHLD;
> + int cpid;
> + char *childStack;
> + char *stack;
> + int childStatus;
> +
> + stack = malloc(getpagesize() * 4);
> + if(!stack) {
> + DEBUG0("Unable to allocate stack");
> + rc = -1;
> + goto check_complete;
> + }
> +
> + childStack = stack + (getpagesize() * 4);
> +
> + cpid = clone(lxcDummyChild, childStack, flags, NULL);
> + if ((0 > cpid) && (EINVAL == errno)) {
> + DEBUG0("clone call returned EINVAL, container support is not
enabled");
> + rc = -1;
haha, I would have expected a checking of /proc or something similar.
That test could still fail, say if the kernel started to disagree on
stack of only 4 pages for example.
If there was insufficient mem it would return ENOMEM. The EINVAL errors are all
related to invalid flags, so this is a reasonable check.
Dan.
--
|: Red Hat, Engineering, Boston -o- http://people.redhat.com/berrange/ :|
|: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|
4:43 p.m.
Daniel Veillard wrote:
On Wed, Mar 19, 2008 at 11:14:59PM -0700, Dave Leskovec wrote:
> This patch adds the lxc_driver source files.
[...]
> +static virDrvOpenStatus lxcOpen(virConnectPtr conn,
> + xmlURIPtr uri,
> + virConnectAuthPtr auth ATTRIBUTE_UNUSED,
> + int flags ATTRIBUTE_UNUSED)
> +{
> + uid_t uid = getuid();
> +
> + /* Check that the user is root */
> + if (0 != uid) {
> + goto declineConnection;
> + }
so it's restricted to root, it's probably fine, as we can go though the
daemon for normal users, ssuming they get authenticated.
Yes it's restricted to root. That could be removed if file capabilities were
set appropriately. I'll look into how feasible that would be.
[...]
> +static int lxcListDomains(virConnectPtr conn, int *ids, int nids)
> +{
> + lxc_driver_t *driver = (lxc_driver_t *)conn->privateData;
> + lxc_vm_t *vm;
> + int numDoms = 0;
> +
> + for (vm = driver->vms; vm && (numDoms < nids); vm = vm->next)
{
> + if (lxcIsActiveVM(vm)) {
> + ids[numDoms] = vm->def->id;
> + numDoms++;
> + }
> + }
> +
> + return numDoms;
> +}
so we can only list domains created by this libvirt instance, right ?
Or I'm missing something, I assume virsh list works but I don't see how.
Well, yes and no. The list of vms is local to the process however all container
configs are stored to file when they're created. So, a later instance of
libvirt (later being after a container is created) will pick up the config file
and know about that container. However, if 2 instances of libvirt are running
and one creates a container, the other won't know about it until it's restarted
or reconnected. This and a few related issues have been sticking in the back of
my mind for a little while. I'm wondering if the solution isn't to have the lxc
driver under libvirtd. That or load and unload the list of vms around every
operation.
Except this bit I don't understand this looks fine to me
> Index: b/src/lxc_driver.h
Looks fine,
thanks !
Daniel
--
Best Regards,
Dave Leskovec
IBM Linux Technology Center
Open Virtualization
3:09 a.m.
On Thu, Mar 20, 2008 at 02:43:28PM -0700, Dave Leskovec wrote:
Daniel Veillard wrote:
> so it's restricted to root, it's probably fine, as we can go though the
> daemon for normal users, ssuming they get authenticated.
Yes it's restricted to root. That could be removed if file capabilities were
set appropriately. I'll look into how feasible that would be.
[...]
> so we can only list domains created by this libvirt instance,
right ?
> Or I'm missing something, I assume virsh list works but I don't see how.
Well, yes and no. The list of vms is local to the process however all container
configs are stored to file when they're created. So, a later instance of
libvirt (later being after a container is created) will pick up the config file
and know about that container. However, if 2 instances of libvirt are running
and one creates a container, the other won't know about it until it's restarted
or reconnected. This and a few related issues have been sticking in the back of
my mind for a little while. I'm wondering if the solution isn't to have the lxc
driver under libvirtd. That or load and unload the list of vms around every
operation.
It's probably a good option, at least as a first step, maybe a separate
daemon may be needed to keep persistance of runtime data like the open
file descriptors (admitedly I didn't yet tried to digest all the technical
description in part 4). Repetitive load/unload is not acceptable I'm afraid,
we should aim at lightweight virtualization for containers, let's not start
multiplicating I/O access on operations, we have enough pain with the xend
side. Another important feature is not loosing state or data (like file
descriptors for console and the like) when libvirtd is stopped and restarted.
That probably mean we should be able to list running process associated
to containers (if possible in a lightweight fashion).
It may take a bit of time to find the right solutions, the good thing
is that we are completely isolated by the API, and as long as the
support for containers is not declared stable, I'm fine changing some
of the implementation details even if it means a bit of pain during
libvirt details. I'm a believer in pushing code out and refining it based
on feedback :-)
Daniel
--
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard | virtualization library http://libvirt.org/
veillard(a)redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
12:05 p.m.
On Thu, Mar 20, 2008 at 02:43:28PM -0700, Dave Leskovec wrote:
Daniel Veillard wrote:
> [...]
>> +static int lxcListDomains(virConnectPtr conn, int *ids, int nids)
>> +{
>> + lxc_driver_t *driver = (lxc_driver_t *)conn->privateData;
>> + lxc_vm_t *vm;
>> + int numDoms = 0;
>> +
>> + for (vm = driver->vms; vm && (numDoms < nids); vm =
vm->next) {
>> + if (lxcIsActiveVM(vm)) {
>> + ids[numDoms] = vm->def->id;
>> + numDoms++;
>> + }
>> + }
>> +
>> + return numDoms;
>> +}
>
> so we can only list domains created by this libvirt instance, right ?
> Or I'm missing something, I assume virsh list works but I don't see how.
Well, yes and no. The list of vms is local to the process however all container
configs are stored to file when they're created. So, a later instance of
libvirt (later being after a container is created) will pick up the config file
and know about that container. However, if 2 instances of libvirt are running
and one creates a container, the other won't know about it until it's restarted
or reconnected. This and a few related issues have been sticking in the back of
my mind for a little while. I'm wondering if the solution isn't to have the lxc
driver under libvirtd. That or load and unload the list of vms around every
operation.
Yes, this is a fundamentally stateful driver, so should run in the context
of the daemon as the QEMU driver does. Loading & unloading the persistent
state in the open/close method is hacking around the capabilities we already
provide for stateful drivers, and it is not safe to have multiple libvirt
clients reading/writing the same config files without any form of locking.
By using the libvirt daemon stateful driver support we avoid the locking
issues, avoid the synchronization problems between multiple clients, and
avoid having to have the fork()d container re-write the config files to
include the PID and so on.
The following patches make the driver properly integrate with the stateful
driver APIs. It also changes the config files to be named by on VM name
instead of UUID, since this is what the QEMU driver does & its more user
friendly. It also adds the CLONE_XXX constants since they have not yet
been added to the libc sched.h file. IMHO we should enable the driver by
default, since it can already probe for availability at runtime. Finally
it also fixes a typo where it wrote 'linuxcontainer' as the domain type
in the config file instead of 'lxc'.
Dan.
Index: src/lxc_conf.c
===================================================================
RCS file: /data/cvs/libvirt/src/lxc_conf.c,v
retrieving revision 1.1
diff -u -p -r1.1 lxc_conf.c
--- src/lxc_conf.c 21 Mar 2008 15:03:37 -0000 1.1
+++ src/lxc_conf.c 21 Mar 2008 16:47:26 -0000
@@ -593,7 +593,6 @@ int lxcSaveVMDef(virConnectPtr conn,
lxc_vm_def_t *def)
{
int rc = -1;
- char uuidstr[VIR_UUID_STRING_BUFLEN];
if (vm->configFile[0] == '\0') {
if ((rc = virFileMakePath(driver->configDir))) {
@@ -603,16 +602,15 @@ int lxcSaveVMDef(virConnectPtr conn,
goto save_complete;
}
- virUUIDFormat(def->uuid, uuidstr);
- if (virFileBuildPath(driver->configDir, uuidstr, ".xml",
+ if (virFileBuildPath(driver->configDir, vm->def->name,
".xml",
vm->configFile, PATH_MAX) < 0) {
lxcError(conn, NULL, VIR_ERR_INTERNAL_ERROR,
_("cannot construct config file path"));
goto save_complete;
}
- strncpy(vm->configFileBase, uuidstr, PATH_MAX);
- strncat(vm->configFileBase, ".xml", PATH_MAX - strlen(uuidstr));
+ strncpy(vm->configFileBase, vm->def->name, PATH_MAX-1);
+ strncat(vm->configFileBase, ".xml", PATH_MAX -
strlen(vm->def->name)-1);
}
@@ -631,7 +629,6 @@ static lxc_vm_t * lxcLoadConfig(lxc_driv
{
lxc_vm_def_t *containerDef;
lxc_vm_t * vm;
- char uuidstr[VIR_UUID_STRING_BUFLEN];
containerDef = lxcParseVMDef(NULL, xmlData, file);
if (NULL == containerDef) {
@@ -639,9 +636,8 @@ static lxc_vm_t * lxcLoadConfig(lxc_driv
return NULL;
}
- virUUIDFormat(containerDef->uuid, uuidstr);
- if (!virFileMatchesNameSuffix(file, uuidstr, ".xml")) {
- DEBUG0("Container uuid does not match config file name");
+ if (!virFileMatchesNameSuffix(file, containerDef->name, ".xml")) {
+ DEBUG0("Container name does not match config file name");
lxcFreeVMDef(containerDef);
return NULL;
}
@@ -662,14 +658,12 @@ static lxc_vm_t * lxcLoadConfig(lxc_driv
return vm;
}
-int lxcLoadDriverConfig(virConnectPtr conn)
+int lxcLoadDriverConfig(lxc_driver_t *driver)
{
- lxc_driver_t *driverPtr = (lxc_driver_t*)conn->privateData;
-
/* Set the container configuration directory */
- driverPtr->configDir = strdup(SYSCONF_DIR "/libvirt/lxc");
- if (NULL == driverPtr->configDir) {
- lxcError(conn, NULL, VIR_ERR_NO_MEMORY, "configDir");
+ driver->configDir = strdup(SYSCONF_DIR "/libvirt/lxc");
+ if (NULL == driver->configDir) {
+ lxcError(NULL, NULL, VIR_ERR_NO_MEMORY, "configDir");
return -1;
}
@@ -702,19 +696,18 @@ load_complete:
return rc;
}
-int lxcLoadContainerInfo(virConnectPtr conn)
+int lxcLoadContainerInfo(lxc_driver_t *driver)
{
int rc = -1;
- lxc_driver_t *driverPtr = (lxc_driver_t*)conn->privateData;
DIR *dir;
struct dirent *dirEntry;
- if (!(dir = opendir(driverPtr->configDir))) {
+ if (!(dir = opendir(driver->configDir))) {
if (ENOENT == errno) {
/* no config dir => no containers */
rc = 0;
} else {
- lxcError(conn, NULL, VIR_ERR_INTERNAL_ERROR,
+ lxcError(NULL, NULL, VIR_ERR_INTERNAL_ERROR,
_("failed to open config directory: %s"),
strerror(errno));
}
@@ -730,7 +723,7 @@ int lxcLoadContainerInfo(virConnectPtr c
continue;
}
- lxcLoadContainerConfigFile(driverPtr, dirEntry->d_name);
+ lxcLoadContainerConfigFile(driver, dirEntry->d_name);
}
closedir(dir);
@@ -758,12 +751,13 @@ char *lxcGenerateXML(virConnectPtr conn,
}
if (lxcIsActiveVM(vm)) {
- if (virBufferVSprintf(buf, "<domain type='linuxcontainer'
id='%d'>\n",
- vm->def->id) < 0) {
+ if (virBufferVSprintf(buf, "<domain type='%s'
id='%d'>\n",
+ LXC_DOMAIN_TYPE, vm->def->id) < 0) {
goto no_memory;
}
} else {
- if (virBufferAddLit(buf, "<domain
type='linuxcontainer'>\n") < 0) {
+ if (virBufferVSprintf(buf, "<domain type='%s'>\n",
+ LXC_DOMAIN_TYPE) < 0) {
goto no_memory;
}
}
@@ -940,3 +934,12 @@ int lxcDeleteConfig(virConnectPtr conn,
#endif /* WITH_LXC */
+
+/*
+ * Local variables:
+ * indent-tabs-mode: nil
+ * c-indent-level: 4
+ * c-basic-offset: 4
+ * tab-width: 4
+ * End:
+ */
Index: src/lxc_conf.h
===================================================================
RCS file: /data/cvs/libvirt/src/lxc_conf.h,v
retrieving revision 1.1
diff -u -p -r1.1 lxc_conf.h
--- src/lxc_conf.h 21 Mar 2008 15:03:37 -0000 1.1
+++ src/lxc_conf.h 21 Mar 2008 16:47:26 -0000
@@ -102,12 +102,12 @@ int lxcSaveVMDef(virConnectPtr conn,
lxc_driver_t *driver,
lxc_vm_t *vm,
lxc_vm_def_t *def);
-int lxcLoadDriverConfig(virConnectPtr conn);
+int lxcLoadDriverConfig(lxc_driver_t *driver);
int lxcSaveConfig(virConnectPtr conn,
lxc_driver_t *driver,
lxc_vm_t *vm,
lxc_vm_def_t *def);
-int lxcLoadContainerInfo(virConnectPtr conn);
+int lxcLoadContainerInfo(lxc_driver_t *driver);
int lxcLoadContainerConfigFile(lxc_driver_t *driver,
const char *file);
lxc_vm_t * lxcAssignVMDef(virConnectPtr conn,
Index: src/lxc_driver.c
===================================================================
RCS file: /data/cvs/libvirt/src/lxc_driver.c,v
retrieving revision 1.1
diff -u -p -r1.1 lxc_driver.c
--- src/lxc_driver.c 21 Mar 2008 15:03:37 -0000 1.1
+++ src/lxc_driver.c 21 Mar 2008 16:47:26 -0000
@@ -41,8 +41,27 @@
#define DEBUG(fmt,...) VIR_DEBUG(__FILE__, fmt, __VA_ARGS__)
#define DEBUG0(msg) VIR_DEBUG(__FILE__, "%s", msg)
-static int lxcStartup(virConnectPtr conn);
-static int lxcShutdown(virConnectPtr conn);
+/*
+ * GLibc headers are behind the kernel, so we define these
+ * constants if they're not present already.
+ */
+
+#ifndef CLONE_NEWPID
+#define CLONE_NEWPID 0x20000000
+#endif
+#ifndef CLONE_NEWUTS
+#define CLONE_NEWUTS 0x04000000
+#endif
+#ifndef CLONE_NEWUSER
+#define CLONE_NEWUSER 0x10000000
+#endif
+#ifndef CLONE_NEWIPC
+#define CLONE_NEWIPC 0x08000000
+#endif
+
+static int lxcStartup(void);
+static int lxcShutdown(void);
+static lxc_driver_t *lxc_driver;
/* Functions */
static int lxcDummyChild( void *argv ATTRIBUTE_UNUSED )
@@ -105,6 +124,9 @@ static virDrvOpenStatus lxcOpen(virConne
goto declineConnection;
}
+ if (lxc_driver == NULL)
+ goto declineConnection;
+
/* Verify uri was specified */
if ((NULL == uri) || (NULL == uri->scheme)) {
goto declineConnection;
@@ -115,15 +137,7 @@ static virDrvOpenStatus lxcOpen(virConne
goto declineConnection;
}
- /* Check that this is a container enabled kernel */
- if(0 != lxcCheckContainerSupport()) {
- goto declineConnection;
- }
-
- /* initialize driver data */
- if (0 > lxcStartup(conn)) {
- goto declineConnection;
- }
+ conn->privateData = lxc_driver;
return VIR_DRV_OPEN_SUCCESS;
@@ -133,7 +147,8 @@ declineConnection:
static int lxcClose(virConnectPtr conn)
{
- return lxcShutdown(conn);
+ conn->privateData = NULL;
+ return 0;
}
static virDomainPtr lxcDomainLookupByID(virConnectPtr conn,
@@ -360,26 +375,29 @@ static char *lxcDomainDumpXML(virDomainP
return lxcGenerateXML(dom->conn, driver, vm, vm->def);
}
-static int lxcStartup(virConnectPtr conn)
+
+static int lxcStartup(void)
{
- lxc_driver_t *driver;
+ lxc_driver = calloc(1, sizeof(lxc_driver_t));
+ if (NULL == lxc_driver) {
+ return -1;
+ }
- driver = calloc(1, sizeof(lxc_driver_t));
- if (NULL == driver) {
+ /* Check that this is a container enabled kernel */
+ if(0 != lxcCheckContainerSupport()) {
return -1;
}
- conn->privateData = driver;
/* Call function to load lxc driver configuration information */
- if (lxcLoadDriverConfig(conn) < 0) {
- lxcShutdown(conn);
+ if (lxcLoadDriverConfig(lxc_driver) < 0) {
+ lxcShutdown();
return -1;
}
/* Call function to load the container configuration files */
- if (lxcLoadContainerInfo(conn) < 0) {
- lxcShutdown(conn);
+ if (lxcLoadContainerInfo(lxc_driver) < 0) {
+ lxcShutdown();
return -1;
}
@@ -392,19 +410,37 @@ static void lxcFreeDriver(lxc_driver_t *
free(driver);
}
-static int lxcShutdown(virConnectPtr conn)
+static int lxcShutdown(void)
{
- lxc_driver_t *driver = (lxc_driver_t *)conn->privateData;
- lxc_vm_t *vms = driver->vms;
+ lxc_vm_t *vms = lxc_driver->vms;
lxcFreeVMs(vms);
- driver->vms = NULL;
- lxcFreeDriver(driver);
- conn->privateData = NULL;
+ lxc_driver->vms = NULL;
+ lxcFreeDriver(lxc_driver);
return 0;
}
+/**
+ * lxcActive:
+ *
+ * Checks if the LXC daemon is active, i.e. has an active domain
+ *
+ * Returns 1 if active, 0 otherwise
+ */
+static int
+lxcActive(void) {
+ /* If we've any active networks or guests, then we
+ * mark this driver as active
+ */
+ if (lxc_driver->nactivevms)
+ return 1;
+
+ /* Otherwise we're happy to deal with a shutdown */
+ return 0;
+}
+
+
/* Function Tables */
static virDriver lxcDriver = {
VIR_DRV_LXC, /* the number virDrvNo */
@@ -466,9 +502,18 @@ static virDriver lxcDriver = {
NULL, /* getFreeMemory */
};
+
+static virStateDriver lxcStateDriver = {
+ lxcStartup,
+ lxcShutdown,
+ NULL, /* reload */
+ lxcActive,
+};
+
int lxcRegister(void)
{
virRegisterDriver(&lxcDriver);
+ virRegisterStateDriver(&lxcStateDriver);
return 0;
}
Index: src/remote_internal.c
===================================================================
RCS file: /data/cvs/libvirt/src/remote_internal.c,v
retrieving revision 1.63
diff -u -p -r1.63 remote_internal.c
--- src/remote_internal.c 17 Mar 2008 17:30:48 -0000 1.63
+++ src/remote_internal.c 21 Mar 2008 16:47:32 -0000
@@ -843,6 +843,12 @@ remoteOpen (virConnectPtr conn,
rflags |= VIR_DRV_OPEN_REMOTE_UNIX;
}
#endif
+#if WITH_LXC
+ if (uri &&
+ uri->scheme && STREQ (uri->scheme, "lxc")) {
+ rflags |= VIR_DRV_OPEN_REMOTE_UNIX;
+ }
+#endif
priv->magic = DEAD;
priv->sock = -1;
--
|: Red Hat, Engineering, Boston -o- http://people.redhat.com/berrange/ :|
|: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|
12:32 p.m.
"Daniel P. Berrange" <berrange(a)redhat.com> wrote:
...
The following patches make the driver properly integrate with the
stateful
driver APIs. It also changes the config files to be named by on VM name
instead of UUID, since this is what the QEMU driver does & its more user
friendly. It also adds the CLONE_XXX constants since they have not yet
been added to the libc sched.h file. IMHO we should enable the driver by
default, since it can already probe for availability at runtime. Finally
it also fixes a typo where it wrote 'linuxcontainer' as the domain type
in the config file instead of 'lxc'.
...
Index: src/lxc_driver.c
===================================================================
RCS file: /data/cvs/libvirt/src/lxc_driver.c,v
...
+static lxc_driver_t *lxc_driver;
Do you really want to add a static variable to the library?
12:36 p.m.
On Fri, Mar 21, 2008 at 06:32:45PM +0100, Jim Meyering wrote:
"Daniel P. Berrange" <berrange(a)redhat.com> wrote:
...
> The following patches make the driver properly integrate with the stateful
> driver APIs. It also changes the config files to be named by on VM name
> instead of UUID, since this is what the QEMU driver does & its more user
> friendly. It also adds the CLONE_XXX constants since they have not yet
> been added to the libc sched.h file. IMHO we should enable the driver by
> default, since it can already probe for availability at runtime. Finally
> it also fixes a typo where it wrote 'linuxcontainer' as the domain type
> in the config file instead of 'lxc'.
...
> Index: src/lxc_driver.c
> ===================================================================
> RCS file: /data/cvs/libvirt/src/lxc_driver.c,v
...
> +static lxc_driver_t *lxc_driver;
Do you really want to add a static variable to the library?
Yes, this turns LXC into a stateful driver that lives in the context of the
daemon only. This is the same way the QEMU / network drivers work.
Regards,
Dan.
--
|: Red Hat, Engineering, Boston -o- http://people.redhat.com/berrange/ :|
|: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|
3:37 p.m.
On Fri, Mar 21, 2008 at 05:05:43PM +0000, Daniel P. Berrange wrote:
The following patches make the driver properly integrate with the
stateful
driver APIs. It also changes the config files to be named by on VM name
instead of UUID, since this is what the QEMU driver does & its more user
friendly. It also adds the CLONE_XXX constants since they have not yet
been added to the libc sched.h file. IMHO we should enable the driver by
default, since it can already probe for availability at runtime. Finally
it also fixes a typo where it wrote 'linuxcontainer' as the domain type
in the config file instead of 'lxc'.
Looks the best way, yes, +1
Daniel
--
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard | virtualization library http://libvirt.org/
veillard(a)redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
4:14 p.m.
Daniel P. Berrange wrote:
The following patches make the driver properly integrate with the
stateful
driver APIs. It also changes the config files to be named by on VM name
instead of UUID, since this is what the QEMU driver does & its more user
friendly. It also adds the CLONE_XXX constants since they have not yet
been added to the libc sched.h file. IMHO we should enable the driver by
default, since it can already probe for availability at runtime. Finally
it also fixes a typo where it wrote 'linuxcontainer' as the domain type
in the config file instead of 'lxc'.
Thanks, this looks great.
Back when I posted the first pass of these patches, you mentioned it should be
defaulted to enabled only on Linux. I'm not an autoconf expert (or intermediate
for that matter). How do we default the enabled parm based on the host os?
--
Best Regards,
Dave Leskovec
IBM Linux Technology Center
Open Virtualization
4:41 a.m.
On Fri, Mar 21, 2008 at 02:14:38PM -0700, Dave Leskovec wrote:
Daniel P. Berrange wrote:
> The following patches make the driver properly integrate with the stateful
> driver APIs. It also changes the config files to be named by on VM name
> instead of UUID, since this is what the QEMU driver does & its more user
> friendly. It also adds the CLONE_XXX constants since they have not yet
> been added to the libc sched.h file. IMHO we should enable the driver by
> default, since it can already probe for availability at runtime. Finally
> it also fixes a typo where it wrote 'linuxcontainer' as the domain type
> in the config file instead of 'lxc'.
Thanks, this looks great.
Back when I posted the first pass of these patches, you mentioned it should be
defaulted to enabled only on Linux. I'm not an autoconf expert (or intermediate
for that matter). How do we default the enabled parm based on the host os?
Hum, maybe in configure.in
if test "$with_lxc" = "yes" ; then
could be tweaked to use the "`uname`" = "Linux" condition too.
But I don't know if the
AC_ARG_WITH(lxc,
[ --with-lxc add Linux Container support (off)],[],[with_lxc=no])
can be enclosed in a similar if test, probably worth trying
Daniel
--
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard | virtualization library http://libvirt.org/
veillard(a)redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
4:34 a.m.
On Fri, Mar 21, 2008 at 05:05:43PM +0000, Daniel P. Berrange wrote:
Yes, this is a fundamentally stateful driver, so should run in the
context
of the daemon as the QEMU driver does. Loading & unloading the persistent
state in the open/close method is hacking around the capabilities we already
provide for stateful drivers, and it is not safe to have multiple libvirt
clients reading/writing the same config files without any form of locking.
By using the libvirt daemon stateful driver support we avoid the locking
issues, avoid the synchronization problems between multiple clients, and
avoid having to have the fork()d container re-write the config files to
include the PID and so on.
The following patches make the driver properly integrate with the stateful
driver APIs. It also changes the config files to be named by on VM name
instead of UUID, since this is what the QEMU driver does & its more user
friendly. It also adds the CLONE_XXX constants since they have not yet
been added to the libc sched.h file. IMHO we should enable the driver by
default, since it can already probe for availability at runtime. Finally
it also fixes a typo where it wrote 'linuxcontainer' as the domain type
in the config file instead of 'lxc'.
I applied the patch in CVs since it was agreed upon and needed for further
developments.
Daniel
--
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard | virtualization library http://libvirt.org/
veillard(a)redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
6086
days inactive
6093
days old
11 comments
4 participants
participants (4)
-
Daniel P. Berrange -
Daniel Veillard -
Dave Leskovec -
Jim Meyering