Daniel P. Berrange berra...@redhat.com wrote:
On Tue, Jan 13, 2009 at 05:45:43PM +, Daniel P. Berrange wrote:
Historically libvirtd was single threaded, serializing all
requests across clients. An recent patch allowed multiple
threads, so multiple clients could run in parallel. A single
On Mon, Jan 19, 2009 at 10:06:47PM -0500, john cooper wrote:
There have been two differnet syntaxes supported in QEMU for caching,
so we need to detect this and switch between them. Originally there
was just
cache=on cache=off (writethrough and no caching)
Now it supports
Small addition to the spec file, basically if on Fedora we should not
activate RHEL-5 Xen specific support, but on RHEL and possibly CentOS
the flag should be passed to %configure.
The test is somehow reversed because it's hard to detect building on
a RHEL platform from the spec file, and that
Daniel P. Berrange berra...@redhat.com wrote:
The libvirtd.conf file has three parameters
max_clients
min_workers
max_workers
When the daemon starts up it spawns min_workers threads. It
accepts connections from upto max_clients. I never implemented
the logic to auto-spawn more
On Tue, Jan 20, 2009 at 08:17:07AM +0100, Guido G?nther wrote:
Hi,
On Mon, Jan 19, 2009 at 01:38:22PM +, Daniel P. Berrange wrote:
/* Got them all, so now open the monitor console */
-ret = qemudOpenMonitor(conn, vm, monitor);
+qemuDriverLock(driver);
+ret =
On Tue, Jan 20, 2009 at 08:18:11AM +0100, Guido G?nther wrote:
On Mon, Jan 19, 2009 at 01:39:51PM +, Daniel P. Berrange wrote:
+vm-stdout_fd = vm-stderr_fd = vm-logfile;
If nothing is actaully using the stdout/err FDs anymore we can just
leave them at -1.
Fixed in the
On Tue, Jan 20, 2009 at 10:12:04AM +0100, Jim Meyering wrote:
} else {
ret = gnutls_record_recv (client-tlssession, data, len);
-if (qemudRegisterClientEvent (server, client, 1) 0)
-qemudDispatchClientFailure (server, client);
-else if (ret = 0)
On Mon, Jan 19, 2009 at 07:41:18PM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
The GNULIB Makefile.am doesn't set any particular compiler warning
flags in the belief that upstream GNULIB developers sort out all
warnings before committing their code. This is
On Tue, Jan 20, 2009 at 08:20:05AM +0100, Guido G?nther wrote:
Hi,
After applying the previous two patches the qemu driver doesn't need
these anymore and they can then be removed from lxc/uml as well.
This is great, ACK.
Daniel
--
|: Red Hat, Engineering, London -o-
On Mon, Jan 19, 2009 at 04:07:33PM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
The 'xm' driver currently keeps all its state in a global static
variables. Not cool, since we access this from all sorts of places
and its hard to guarentee thread safety. So we move
On Mon, Jan 19, 2009 at 07:20:07PM -0500, Mark Wagner wrote:
Daniel P. Berrange wrote:
On Mon, Jan 19, 2009 at 01:53:48PM +0100, Daniel Schwager wrote:
Hi Daniel,
Charles Duffy (thanks Charles !) told me how to fix libvirt for using
the
new -incoming API - yust replace stdio with exec:cat
Hello,
I want to use PXE boot in kvm's virtual machines, but I've not found, how to
configure libvirt for starting dhcp server with correct options for this.
Now i have libvirt-0.5.1-2.fc10.x86_64 and virt-manager-0.6.0-5.fc10.x86_64,
I've created VM using virt manager, with specifying PXE boot
On Fri, Jan 16, 2009 at 01:35:37PM +0100, Daniel Veillard wrote:
On Thu, Jan 15, 2009 at 05:18:40PM +, Richard W.M. Jones wrote:
On Tue, Jan 13, 2009 at 05:43:08PM +, Daniel P. Berrange wrote:
With the domain events code, the callbacks triggered upon events get given
a
Daniel P. Berrange berra...@redhat.com wrote:
Now that gnulib's rand module is imported, we have a decent
quality random number generator that's portable. We don't
want to mess with the apps state, so in virInitialize() we
explicitly initialize our own private random nubmer generator
state
Daniel P. Berrange berra...@redhat.com wrote:
The configure output looked untidy, so this aligns things
correctly.
configure.in | 20 ++--
1 file changed, 10 insertions(+), 10 deletions(-)
+1, of course
--
Libvir-list mailing list
Libvir-list@redhat.com
On Tue, Jan 20, 2009 at 11:18:40AM +, Daniel P. Berrange wrote:
Unfortunately this isn't sufficient to make use of the dom/net
fields practical. If we hold onto a reference when setting the
error object, then we have a possible indefinite memory leak,
unless the app calling libvirt always
On Tue, Jan 13, 2009 at 05:40:18PM +, Daniel P. Berrange wrote:
Adds locking in the remote driver. This obtains the lock on every single
API method. This is quite poor, not giving any parallelism - that will
be added in 2 patch's time, when we drop the lock while waiting on I/O
On Thu, Jan 15, 2009 at 03:19:39PM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
When going into the call() method, we temporarily disable the
event wakup callback. The call() method is already processing
any async messages which arrive, so letting the main event
On Fri, Jan 16, 2009 at 12:11:16PM +, Daniel P. Berrange wrote:
@@ -114,6 +164,11 @@ struct private_data {
virDomainEventQueuePtr domainEvents;
/* Timer for flushing domainEvents queue */
int eventFlushTimer;
+
+/* List of threads currently doing dispatch
On Mon, Jan 19, 2009 at 09:33:24AM +0100, Jim Meyering wrote:
I updated the list of error function names
in Makefile.maint, and exposed 3 more unmarked diagnostics.
Going through the old/new lists, I also found two
now-unused (#define'd) functions.
I originally did this on top of Dan's
On Fri, Jan 16, 2009 at 10:01:01PM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
The virGetLastError() and virConnGetLastError() methods are not
even remotely thread safe, and the virCopyLastError/virConnCopyLastError
methods don't help in this goal, being open to
On Fri, Jan 16, 2009 at 01:35:37PM +0100, Daniel Veillard wrote:
On Thu, Jan 15, 2009 at 05:18:40PM +, Richard W.M. Jones wrote:
On Tue, Jan 13, 2009 at 05:43:08PM +, Daniel P. Berrange wrote:
With the domain events code, the callbacks triggered upon events get given
a
On Fri, Jan 16, 2009 at 04:43:52PM +0100, Daniel Veillard wrote:
On Fri, Jan 16, 2009 at 12:24:36PM +, Richard W.M. Jones wrote:
On Tue, Jan 13, 2009 at 05:47:49PM +, Daniel P. Berrange wrote:
The configure output looked untidy, so this aligns things
correctly.
Daniel P. Berrange berra...@redhat.com wrote:
This adds support for the domain events in the test driver. Code
is following the same pattern as the impl in the QEMU driver.
test.c | 223
+++--
1 file changed, 217 insertions(+), 6
Daniel P. Berrange berra...@redhat.com wrote:
On Fri, Jan 16, 2009 at 12:11:16PM +, Daniel P. Berrange wrote:
@@ -114,6 +164,11 @@ struct private_data {
virDomainEventQueuePtr domainEvents;
/* Timer for flushing domainEvents queue */
int eventFlushTimer;
+
+
On Mon, Jan 19, 2009 at 10:05:03PM +, John Levon wrote:
On Mon, Jan 19, 2009 at 04:32:44PM +0100, Daniel Veillard wrote:
that this code would fail except in the first time the daemon is
launched because mkdir /var/run/libvirt will return -1 and errno EEXIST
in all following
Hi,
i would like to write a small script
- bringing up my qemu-kvm interface and
- connecting it to a bridge
using the following XML domain-syntax :
interface type='ethernet'
script path='/etc/kvm/qemu-ifup'/
/interface
The
On Tue, Jan 20, 2009 at 03:21:51PM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
Here is an update with those suggested renames bug fixes in it.
It also addresses the error reporting issue mentioned in
On Tue, Jan 20, 2009 at 04:12:06PM +0100, Daniel Schwager wrote:
Hi,
i would like to write a small script
- bringing up my qemu-kvm interface and
- connecting it to a bridge
using the following XML domain-syntax :
interface type='ethernet'
script
Daniel P. Berrange berra...@redhat.com wrote:
...
Nope, neither this patch, nor the original version uses the function
virReportSystemError() since that is only added by a later patch in
the series. There's no regression there.
Oh, good.
Then go for it.
--
Libvir-list mailing list
On Tue, Jan 20, 2009 at 12:46:11PM +, Daniel P. Berrange wrote:
On Mon, Jan 19, 2009 at 11:40:56AM +0100, Jim Meyering wrote:
diff --git a/src/test.c b/src/test.c
...
@@ -336,7 +339,7 @@ static int testOpenFromFile(virConnectPt
virDomainObjPtr dom;
testConnPtr
On Tue, Jan 20, 2009 at 04:44:22PM +0100, Daniel Schwager wrote:
Hi Daniel,
thx for the fast answer.
QEMU is responsible for passing arguments to the interface scripts,
so its out of libvirt's control. QEMU passes a single paramter to
the script, which is the name of the TAP device
On Tue, Jan 20, 2009 at 10:21:44AM +, Daniel P. Berrange wrote:
On Tue, Jan 20, 2009 at 08:17:07AM +0100, Guido G?nther wrote:
Hi,
On Mon, Jan 19, 2009 at 01:38:22PM +, Daniel P. Berrange wrote:
/* Got them all, so now open the monitor console */
-ret =
On Tue, Jan 20, 2009 at 10:22:04AM +, Daniel P. Berrange wrote:
On Tue, Jan 20, 2009 at 08:18:11AM +0100, Guido G?nther wrote:
On Mon, Jan 19, 2009 at 01:39:51PM +, Daniel P. Berrange wrote:
+vm-stdout_fd = vm-stderr_fd = vm-logfile;
If nothing is actaully using the
On Tue, Jan 20, 2009 at 10:18:26AM +, Daniel P. Berrange wrote:
On Tue, Jan 20, 2009 at 08:20:05AM +0100, Guido G?nther wrote:
Hi,
After applying the previous two patches the qemu driver doesn't need
these anymore and they can then be removed from lxc/uml as well.
This is great, ACK.
Daniel P. Berrange berra...@redhat.com wrote:
...
Incidentally, virFileMakePath will have to go, eventually. Not only
is it recursive in the number of components of the name being created,
but allocating PATH_MAX for each stack frame is very wasteful, and might
make it easy to blow the
On Tue, Jan 20, 2009 at 08:17:07AM +0100, Guido Günther wrote:
+if (lseek(fd, pos, SEEK_SET) 0) {
+qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
+ _(Unable to seek to %lld in %s: %s),
+ pos, logfile, strerror(errno));
On Tue, Jan 20, 2009 at 11:15:40AM +0100, Daniel Veillard wrote:
Small addition to the spec file, basically if on Fedora we should not
activate RHEL-5 Xen specific support, but on RHEL and possibly CentOS
the flag should be passed to %configure.
The test is somehow reversed because it's
On Tue, Jan 20, 2009 at 05:16:29PM +, Daniel P. Berrange wrote:
On Tue, Jan 20, 2009 at 11:15:40AM +0100, Daniel Veillard wrote:
Small addition to the spec file, basically if on Fedora we should not
activate RHEL-5 Xen specific support, but on RHEL and possibly CentOS
the flag should
On Tue, Jan 20, 2009 at 10:27:56AM +, Daniel P. Berrange wrote:
The above is dead code, since the condition can never be true.
It should be testing ret 0, not ret == -1.
Also, the 2nd should be ||.
if (ret 0 (ret == GNUTLS_E_AGAIN ||
On Mon, Jan 19, 2009 at 10:56:18PM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
On Tue, Jan 13, 2009 at 05:45:43PM +, Daniel P. Berrange wrote:
Historically libvirtd was single threaded, serializing all
requests across clients. An recent patch allowed
On Tue, Jan 20, 2009 at 11:21:03AM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
The libvirtd.conf file has three parameters
max_clients
min_workers
max_workers
When the daemon starts up it spawns min_workers threads. It
accepts connections from
On Mon, Jan 19, 2009 at 03:14:19PM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
The 'inotify' driver currently keeps all its state in a global static
variables. Not so great since we access this from all sorts of places
and its hard to guarentee thread safety.
David Lutterkort wrote:
For certain applications, we want libvirt to be able to configure host
network interfaces in a variety of ways; currently, we are most
interested in teaching libvirt how to set up ordinary ethernet
interfaces, bridges, bonding and vlan's.
Below is a high-level
Daniel P. Berrange berra...@redhat.com wrote:
...
Finally I've also made use of ssize_t as suggested, this also
made me notice one place where we didn't propagate the EAGAIN
case of ret=0 correctly.
I compared our two patched versions and it all looks fine.
It's good that you made those
Daniel P. Berrange berra...@redhat.com wrote:
On Tue, Jan 20, 2009 at 11:21:03AM +0100, Jim Meyering wrote:
...
It'd be nice to make all of these n* members unsigned,
assuming that they should never go negative.
I've not changed this to unsigned, because again we compare it
in various places
# HG changeset patch
# User john.le...@sun.com
# Date 1232478815 28800
# Node ID 37a6a671e38d7f4449d3b47bc5cf06c19362bc41
# Parent 15bfba92a98744f8c554757ed608c3bf89fa38aa
Fix openlog() ident usage
It cannot be freed, as it's used by reference in the syslog code, at
least on Solaris.
# HG changeset patch
# User john.le...@sun.com
# Date 1232478815 28800
# Node ID 9c7ef175f3a797ecc3ffa6b1fed5a27d1814838c
# Parent ce76623e857f5bb2cf3af8414335f891fd7902b7
Solaris least privilege support
On Solaris dom0, virtd runs as a privilege barrier: all libvirt
connections are routed
# HG changeset patch
# User john.le...@sun.com
# Date 1232478815 28800
# Node ID a9c75296e1d6e09947a231fa9864d67a543979eb
# Parent 37a6a671e38d7f4449d3b47bc5cf06c19362bc41
Set syslog logging priority based on LIBVIRT_DEBUG
Currently, qemud in daemonized is dropping all messages below level 3,
Daniel Veillard veill...@redhat.com wrote:
On Mon, Jan 19, 2009 at 09:33:24AM +0100, Jim Meyering wrote:
I updated the list of error function names
in Makefile.maint, and exposed 3 more unmarked diagnostics.
Going through the old/new lists, I also found two
now-unused (#define'd) functions.
On Tue, Jan 20, 2009 at 11:14:03AM -0800, john.le...@sun.com wrote:
# HG changeset patch
# User john.le...@sun.com
# Date 1232478815 28800
# Node ID 37a6a671e38d7f4449d3b47bc5cf06c19362bc41
# Parent 15bfba92a98744f8c554757ed608c3bf89fa38aa
Fix openlog() ident usage
It cannot be freed, as
On Tue, Jan 20, 2009 at 11:14:35AM -0800, john.le...@sun.com wrote:
# HG changeset patch
# User john.le...@sun.com
# Date 1232478815 28800
# Node ID a9c75296e1d6e09947a231fa9864d67a543979eb
# Parent 37a6a671e38d7f4449d3b47bc5cf06c19362bc41
Set syslog logging priority based on LIBVIRT_DEBUG
On Tue, Jan 20, 2009 at 08:36:27PM +0100, Daniel Veillard wrote:
On Tue, Jan 20, 2009 at 11:14:03AM -0800, john.le...@sun.com wrote:
It cannot be freed, as it's used by reference in the syslog code, at
least on Solaris.
The requirement is apparently not present on Linux, sorry :-)
I
On Tue, Jan 20, 2009 at 09:20:24PM +, Daniel P. Berrange wrote:
On Tue, Jan 20, 2009 at 10:18:31PM +0100, Daniel Veillard wrote:
On Tue, Jan 20, 2009 at 08:36:27PM +0100, Daniel Veillard wrote:
On Tue, Jan 20, 2009 at 11:14:03AM -0800, john.le...@sun.com wrote:
It cannot be freed, as
On Tue, Jan 20, 2009 at 12:22:46PM +0100, Jim Meyering wrote:
Daniel P. Berrange berra...@redhat.com wrote:
Now that gnulib's rand module is imported, we have a decent
quality random number generator that's portable. We don't
want to mess with the apps state, so in virInitialize() we
On Tue, 2009-01-20 at 11:35 -0700, Jim Fehlig wrote:
David Lutterkort wrote:
For certain applications, we want libvirt to be able to configure host
network interfaces in a variety of ways; currently, we are most
interested in teaching libvirt how to set up ordinary ethernet
interfaces,
The domain XML has long supported ability to set the disk driver type
but we only used it in Xen so far.
This patch adds support for QEMU usage. In this usage the driver name
is always going to be 'qemu' since all backends are in QEMU, no alternate
kernelspace / helper daemons as with Xen. The
This is a follow up on Miloslav's proposal to add copy on write
support to the storage APIs, changing the XML to that described
here:
http://www.redhat.com/archives/libvir-list/2009-January/msg00231.html
In addition to the original QCOW/VMDK support, I have done an impl which
can extract the
This patch adds support for using the monitor interface to set the VNC
password
(qemu) change vnc password
Password:
A minor tricky thing is that we can't just send the command and password
all in one go, we must wait for the 'Password' prompt before sending the
password.
When
On Tue, Jan 20, 2009 at 11:08:56PM +, Daniel P. Berrange wrote:
[..snip..]
+static int
+qemudInitPasswords(virConnectPtr conn,
+ virDomainObjPtr vm) {
+char *info = NULL;
+
+/*
+ * NB: Might have more passwords to set in the future. eg a qcow
+ * disk
# HG changeset patch
# User john.le...@sun.com
# Date 1232501231 28800
# Node ID 12e2e9f6eb99f86c2ab20f1e6a7b7a77418e66db
# Parent 39479c2678f822349169112e31e6061af3289548
Fix remote driver crash
Ensure -watch is initialized so it's not used before an
updateHandleImpl is registered.
Daniel P. Berrange berra...@redhat.com wrote:
How about this then
Looks fine. Thanks!
BTW, what's good practice for actually initializing a seed ?
time() ?
time() ^ getpid() ?
anything better ?
You could mix in some more get*id numbers.
If we really care, using gnulib's gethrxtime
62 matches
Mail list logo