Addressed comment from v6 -> v5
Marcelo:
* Support other APPs to operate /sys/fs/resctrl at same time
Libvirt will scan /sys/fs/resctrl again before doing cache allocation.
patch 10 will address this.
Addressed comment from v4 -> v5:
Marcelo:
* Several typos
virResCtrlSetCacheBanks: Set cache banks of a libvirt domain. It will
create new resource domain under `/sys/fs/resctrl` and fill the
schemata according the cache banks configration.
virResCtrlUpdate: Update the schemata after libvirt domain destroy.
Signed-off-by: Eli Qiao
This patch adds new xml element to support cache tune as:
...
...
id: any non-minus number
host_id: reference of the host's cache banks id, it's from capabilities
type: cache bank type
size: should be multiples of the min_size of the bank on host.
vcpus: cache allocation on vcpu set, if
l3data and l3code type of cache banks should be configured pairs.
Signed-off-by: Eli Qiao
---
src/conf/domain_conf.c | 19 +++
src/util/virresctrl.c | 1 -
src/util/virresctrl.h | 2 ++
3 files changed, 21 insertions(+), 1 deletion(-)
diff --git
This patch support l3 cache allocation compatible mode if cdp enabled
on host.
In this case l3code/l3data has same schemata.
Signed-off-by: Eli Qiao
---
src/conf/domain_conf.c | 15 +--
src/util/virresctrl.c | 10 ++
2 files changed, 23
Other application may touch resctrl while libvirt's running, scan
resctrl again before allocating cache information to the newly created
VM.
Signed-off-by: Eli Qiao
---
src/util/virresctrl.c | 48 +++-
1 file changed, 47
This patch expose cache information to host's capabilites xml.
For l3 cache allocation
For l3 cache allocation supported cdp(seperate data/code):
RFC on mailing list.
https://www.redhat.com/archives/libvir-list/2017-January/msg00644.html
Set cache banks while booting a new domain.
Signed-off-by: Eli Qiao
---
src/qemu/qemu_driver.c | 6 --
src/qemu/qemu_process.c | 53 +
2 files changed, 57 insertions(+), 2 deletions(-)
diff --git
Enable l3code/l3data while doing cache tune.
l3code/l3data should use a continus cbm in their seperated schemata and
the cache size are shared between them, so we need to deal them
differently with l3 cache.
This should enable cdp feature while mounting /sys/fs/resctrl, eg:
mount -t resctrl
The internal struct list domainall is a list which are resctral domain
status shared by all VMs, especiall the default domain, each VM should
access it concomitantly. Ues a mutex to control it.
Each bank's cache_left field is also a global shared resource we need
to be care, add a mutex for each
This patch adds some utils struct and functions to expose resctrl
information.
virResCtrlAvailable: If resctrl interface exist on host
virResCtrlGet: get specify type resource contral information
virResCtrlInit: initialize resctrl struct from the host's sys fs.
resctrlall[]: an array to maintain
--
Eli Qiao
Sent with Sparrow (http://www.sparrowmailapp.com/?sig)
On Wednesday, 15 February 2017 at 1:14 AM, Marcelo Tosatti wrote:
> On Tue, Feb 14, 2017 at 09:37:02AM +0100, Martin Kletzander wrote:
> > On Mon, Feb 13, 2017 at 04:09:13PM -0200, Marcelo Tosatti wrote:
> > > On Mon, Feb
On 02/14/2017 04:44 AM, Joao Martins wrote:
On 02/14/2017 03:13 AM, Jim Fehlig wrote:
On 02/07/2017 05:35 PM, Joao Martins wrote:
This allows us to reuse a single function for both tunnelled and
non-tunnelled variants.
Signed-off-by: Joao Martins
---
New in v2
---
In PinEmulator() and PinIOThread() there is an identical code that
converts []bool into a bitmask. It calculates the location in the
bitmask and then sets it always to 1, instead of looking at the
actual bool value.
---
domain.go | 16 ++--
1 file changed, 10 insertions(+), 6
Please ignore this patch, I failed to see that in 2 out of 4 locations we
already are looking at the value. I'm sending a better patch instead.
On Tue, Feb 14, 2017 at 4:48 PM, Leonid Podolny wrote:
> In PinVcpu(), PinVcpuFlags(), PinEmulator() and PinIOThread() there is
>
In PinVcpu(), PinVcpuFlags(), PinEmulator() and PinIOThread() there is
an almost identical code that converts []bool into a bitmask. It
calculates the location in the bitmask and then sets it always to 1,
instead of looking at the actual bool value.
---
domain.go | 16
1 file
From: Marc-André Lureau
Add a new attribute 'rendernode' to spice element.
Give it to QEMU if qemu supports it (queued for 2.9).
Signed-off-by: Marc-André Lureau
---
docs/formatdomain.html.in | 7 +-
From: Marc-André Lureau
This should have been added with c4a4603de (or 0bdefd9b04).
Signed-off-by: Marc-André Lureau
---
src/conf/node_device_conf.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/src/conf/node_device_conf.c
From: Marc-André Lureau
Add new top-level element, that list the associated
/dev files. Distinguish the main /dev name from symlinks with a 'type'
attribute of value 'dev' or 'symlink'.
Update a test to check XML schema, and actually add it to the test list
since
From: Marc-André Lureau
This is a cosmetic change, shouldn't change XML parsing, and doesn't
break any test.
Signed-off-by: Marc-André Lureau
---
src/conf/node_device_conf.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff
From: Marc-André Lureau
Add a new 'drm' capability for Direct Rendering Manager (DRM) devices,
providing device type information.
Teach the udev backend to populate those devices.
Signed-off-by: Marc-André Lureau
---
Hi,
On Tue, 14 Feb 2017 15:58:45 +
"Richard W.M. Jones" wrote:
> The patches compile.
>
> I looked at both commits and they at least superficially seem
> sensible. I'm not intimately familiar enough with the original code
> to review this fully.
>
> However I want to
From: Marc-André Lureau
Hi,
The following series implements DRM nodedev, exposes the devices /dev
path and symlinks, and allows to configure qemu to use a specific DRM
rendernode. (also included a small unrelated cosmetic change)
thanks!
Marc-André Lureau (5):
On Tue, 2017-02-14 at 17:39 +0100, Michal Privoznik wrote:
> > @@ -645,6 +647,8 @@ Requires: xz
> > %if 0%{?fedora} >= 24
> > Requires: systemd-container
> > %endif
> > +# For managing ACLs
> > +Requires: libacl
>
> This shouldn't be needed. If we are building with libacl-devel, we are
From: Derbyshev Dmitry
This makes it possible to avoid allocations in
virEventPollMakePollFDs, as well as generally reduces time spent in
kernel if handles remain the same, which should generally be true.
---
configure.ac | 28
From: Derbyshev Dmitry
To introduce new implementation (ex. with epoll) some code should be
substituted. This patch divides vireventpoll into common and
implementation-specific parts.
---
src/Makefile.am | 4 +-
From: Derbyshev Dmitry
Provides about 20% boost on local machine with 35 vms.
virEventPollDispatchHandles can also be split to pass cb via epoll
data field.
Should start sending as PATCH instead?
Changes since v1:
* ifdef supstituded by 2 .c files with
On Tue, 2017-02-14 at 16:59 +, Daniel P. Berrange wrote:
> > [1] I know I did while trying to figure this bug out ;)
>
> If we disabled namespace support when libacl is missing at
> build time you would have noticed quite quickly that you
> weren't using namespaces.
Not sure how quickly
On Mon, Feb 13, 2017 at 03:46:38PM +0800, Eli Qiao wrote:
> Addressed comment from v4 -> v5:
>
> Marcelo:
>* Several typos
>* Use flock instead of virFileLock
>
> Addressed comment from v3 -> v4:
>
> Daniel & Marcelo:
>* Added concurrence support
>
> Addressed comment
On Tue, Feb 14, 2017 at 09:37:02AM +0100, Martin Kletzander wrote:
> On Mon, Feb 13, 2017 at 04:09:13PM -0200, Marcelo Tosatti wrote:
> >On Mon, Feb 13, 2017 at 03:42:04PM +0800, Eli Qiao wrote:
> >>> > L3data:0=0xff00;...
> >>> > L3code:0=0xff00;...
> >>> >
> >>> > * Would you please help to test
On Tue, Feb 14, 2017 at 04:59:52PM +0800, Eli Qiao wrote:
>
>
> --
> Eli Qiao
> Sent with Sparrow (http://www.sparrowmailapp.com/?sig)
>
>
> On Tuesday, 14 February 2017 at 4:37 PM, Martin Kletzander wrote:
>
> > On Mon, Feb 13, 2017 at 04:09:13PM -0200, Marcelo Tosatti wrote:
> > > On Mon,
On Tue, Feb 14, 2017 at 05:47:27PM +0100, Andrea Bolognani wrote:
> On Tue, 2017-02-14 at 16:20 +, Daniel P. Berrange wrote:
> > > On the other hand, we really only care about having the ACL
> > > APIs when we are isolating QEMU, which only happens of Linux
> > > due to the namespaces
On Tue, 14 Feb 2017 16:50:14 +0100
Martin Polednik wrote:
> On 07/02/17 12:29 -0700, Alex Williamson wrote:
> >On Tue, 7 Feb 2017 17:26:51 +0100
> >Erik Skultety wrote:
> >
> >> On Mon, Feb 06, 2017 at 09:33:14AM -0700, Alex Williamson wrote:
> >>
On Tue, 2017-02-14 at 16:20 +, Daniel P. Berrange wrote:
> > On the other hand, we really only care about having the ACL
> > APIs when we are isolating QEMU, which only happens of Linux
> > due to the namespaces requirement... So maybe we could have
> > it as a strict requirement on Linux
On 02/14/2017 04:33 PM, Andrea Bolognani wrote:
> Now that they're required by configure, we should make sure
> they are available when building the rpm and when installing it.
> ---
> libvirt.spec.in | 4
> 1 file changed, 4 insertions(+)
>
> diff --git a/libvirt.spec.in b/libvirt.spec.in
On 02/14/2017 04:33 PM, Andrea Bolognani wrote:
> mknod() is affected my the current umask, so we're not
> guaranteed the newly-created device node will have the
> right permissions.
>
> Call chmod(), which is not affected by the current umask,
> immediately afterwards to solve the issue.
> ---
>
On 02/14/2017 04:33 PM, Andrea Bolognani wrote:
> ---
> configure.ac | 12 +++-
> m4/virt-acl.m4 | 35 +++
> 2 files changed, 38 insertions(+), 9 deletions(-)
> create mode 100644 m4/virt-acl.m4
ACK
Michal
--
libvir-list mailing list
On Tue, Feb 14, 2017 at 05:14:33PM +0100, Andrea Bolognani wrote:
> On Tue, 2017-02-14 at 16:00 +, Daniel P. Berrange wrote:
> > Your platform has libacl available so it is not difficult to fix that
> > by building with libacl support. We print out the configure summary
> > precisely so users
On Tue, 2017-02-14 at 16:00 +, Daniel P. Berrange wrote:
> Your platform has libacl available so it is not difficult to fix that
> by building with libacl support. We print out the configure summary
> precisely so users can see if there's any libraries they forgot to
> install which might be
On Tue, Feb 14, 2017 at 04:55:05PM +0100, Michal Privoznik wrote:
> On 02/14/2017 04:40 PM, Daniel P. Berrange wrote:
> > On Tue, Feb 14, 2017 at 04:33:09PM +0100, Andrea Bolognani wrote:
> >> When we're building a private /dev for the isolated QEMU
> >> process, we want to be able to replicate
The patches compile.
I looked at both commits and they at least superficially seem
sensible. I'm not intimately familiar enough with the original code
to review this fully.
However I want to try to test this using libguestfs. I believe the
following test case should be sufficient:
$ cd
On 07/02/17 12:29 -0700, Alex Williamson wrote:
On Tue, 7 Feb 2017 17:26:51 +0100
Erik Skultety wrote:
On Mon, Feb 06, 2017 at 09:33:14AM -0700, Alex Williamson wrote:
> On Mon, 6 Feb 2017 13:19:42 +0100
> Erik Skultety wrote:
>
> > Finally. It's
On 02/14/2017 04:40 PM, Daniel P. Berrange wrote:
> On Tue, Feb 14, 2017 at 04:33:09PM +0100, Andrea Bolognani wrote:
>> When we're building a private /dev for the isolated QEMU
>> process, we want to be able to replicate the contents of
>> the original /dev as closely as possible, including ACLs.
On 02/03/2017 06:32 PM, Jim Fehlig wrote:
> If the apparmor security driver is loaded/enabled and domain config
> contains a element whose type attribute is not 'apparmor',
> starting the domain fails when attempting to label resources such
> as tap FDs.
>
> Many of the apparmor driver entry
On 02/03/2017 06:32 PM, Jim Fehlig wrote:
> Like other callers of reload_profile, don't overwrite errors in
> AppArmorSetSecurityHostdevLabelHelper.
>
> Signed-off-by: Jim Fehlig
> ---
> src/security/security_apparmor.c | 14 +-
> 1 file changed, 1 insertion(+), 13
On Tue, Feb 14, 2017 at 04:33:09PM +0100, Andrea Bolognani wrote:
> When we're building a private /dev for the isolated QEMU
> process, we want to be able to replicate the contents of
> the original /dev as closely as possible, including ACLs.
>
> To ensure that's always possible, make ACL
mknod() is affected my the current umask, so we're not
guaranteed the newly-created device node will have the
right permissions.
Call chmod(), which is not affected by the current umask,
immediately afterwards to solve the issue.
---
src/qemu/qemu_domain.c | 10 ++
1 file changed, 10
Now that they're required by configure, we should make sure
they are available when building the rpm and when installing it.
---
libvirt.spec.in | 4
1 file changed, 4 insertions(+)
diff --git a/libvirt.spec.in b/libvirt.spec.in
index e8c272b..ca569e3 100644
--- a/libvirt.spec.in
+++
---
configure.ac | 12 +++-
m4/virt-acl.m4 | 35 +++
2 files changed, 38 insertions(+), 9 deletions(-)
create mode 100644 m4/virt-acl.m4
diff --git a/configure.ac b/configure.ac
index a995a05..ed49daa 100644
--- a/configure.ac
+++ b/configure.ac
@@
When we're building a private /dev for the isolated QEMU
process, we want to be able to replicate the contents of
the original /dev as closely as possible, including ACLs.
To ensure that's always possible, make ACL support mandatory
when the QEMU driver is enabled.
Resolves:
Several heads were banged against several walls while trying
to figure this out. No animals were harmed in the writing of
this series, though.
Andrea Bolognani (4):
configure: Move ACL checks to a separate file
configure: Make ACL mandatory when building the QEMU driver
spec: Require libacl
On Tue, Feb 14, 2017 at 8:47 PM, Michal Privoznik
wrote:
> On 02/03/2017 10:25 AM, Nitesh Konkar wrote:
> > Currently disk names do not follow the
> > (regex) /^[fhv]d[a-z]+[0-9]*$/ completely
> > and hence one can assign disk names like
> > vd2 etc. This patch ensures that
On 02/14/2017 01:46 PM, Ján Tomko wrote:
> To make sure bit 'b' fits into the bitmap, we need to allocate b+1
> bits, since we number from 0.
>
> Adjust the bitmap test to set a bit at a multiple of 16.
> That way the test fails without this fix, because the VIR_REALLOC
> call clears the newly
On 02/03/2017 10:25 AM, Nitesh Konkar wrote:
> Currently disk names do not follow the
> (regex) /^[fhv]d[a-z]+[0-9]*$/ completely
> and hence one can assign disk names like
> vd2 etc. This patch ensures that the
> disk names follow the regex mentioned.
>
> Signed-off-by: Nitesh Konkar
On 02/14/2017 12:38 PM, Jiri Denemark wrote:
> Commit 2a8d40f4ec refactored qemuMonitorJSONGetCPUx86Data and replaced
> virJSONValueObjectGet(reply, "return") with virJSONValueObjectGetArray.
> While the former is guaranteed to always return non-NULL pointer the
> latter may return NULL if the
Now that we have some qemuSecurity wrappers over
virSecurityManager APIs, lets make sure everybody sticks with
them. We have them for a reason and calling virSecurityManager
API directly instead of wrapper may lead into accidentally
labelling a file on the host instead of namespace.
To make sure bit 'b' fits into the bitmap, we need to allocate b+1
bits, since we number from 0.
Adjust the bitmap test to set a bit at a multiple of 16.
That way the test fails without this fix, because the VIR_REALLOC
call clears the newly added memory even if the original pointer
has not
On Mon, Nov 21, 2016 at 05:00:48PM +0100, Jiri Denemark wrote:
ARAT feature was first introduced in QEMU 2.4.0, which means host-model
CPU mode is unusable with QEMU < 2.4.0 on any host CPU which supports
ARAT. Let's not include this feature in host-model CPUs unless a user
explicitly asks for
On Mon, Nov 14, 2016 at 08:19:15AM -0500, Matt Broadstone wrote:
Presently domain events are emitted for the "agent lifecycle" which
is a specialization on virtio-serial events specific to the channel
named "org.qemu.guest_agent.0". This patch adds a generic event for
"channel lifecycles", which
On Tue, 2017-02-14 at 11:37 +0100, Michal Privoznik wrote:
> > @@ -7040,6 +7040,7 @@ qemuDomainCreateDeviceRecursive(const char *device,
> > #ifdef WITH_SELINUX
> > char *tcon = NULL;
> > #endif
> > +mode_t oldUmask = umask((mode_t) 0);
> >
> > if (!ttl) {
> >
On 02/14/2017 03:34 AM, Jim Fehlig wrote:
> On 02/07/2017 05:35 PM, Joao Martins wrote:
>> From: Bob Liu
>>
>> Tunnelled migration doesn't require any extra network connections beside
>> the libvirt daemon. It's capable of strong encryption and the default
>> option of
On 02/14/2017 03:13 AM, Jim Fehlig wrote:
> On 02/07/2017 05:35 PM, Joao Martins wrote:
>> This allows us to reuse a single function for both tunnelled and
>> non-tunnelled variants.
>>
>> Signed-off-by: Joao Martins
>> ---
>> New in v2
>> ---
>>
Commit 2a8d40f4ec refactored qemuMonitorJSONGetCPUx86Data and replaced
virJSONValueObjectGet(reply, "return") with virJSONValueObjectGetArray.
While the former is guaranteed to always return non-NULL pointer the
latter may return NULL if the returned JSON object is not an array.
Signed-off-by:
polite ping.
On Fri, Feb 3, 2017 at 2:55 PM, Nitesh Konkar <
niteshkonkar.libv...@gmail.com> wrote:
> Currently disk names do not follow the
> (regex) /^[fhv]d[a-z]+[0-9]*$/ completely
> and hence one can assign disk names like
> vd2 etc. This patch ensures that the
> disk names follow the regex
On 02/13/2017 09:18 PM, Andrea Bolognani wrote:
> When we populate the private /dev that's going to be used by
> an isolated QEMU process, we take care all metadata matches
> what's in the top-level namespace: in particular, we copy the
> file permissions directly.
>
> However, since the
On Mon, Feb 13, 2017 at 11:49:06PM +0100, Tomáš Golembiovský wrote:
Add suppression for memory leak in bash observerd with bash 4.4.011 on
Arch Linux.
Signed-off-by: Tomáš Golembiovský
---
tests/.valgrind.supp | 9 +
1 file changed, 9 insertions(+)
diff --git
--
Eli Qiao
Sent with Sparrow (http://www.sparrowmailapp.com/?sig)
On Tuesday, 14 February 2017 at 4:37 PM, Martin Kletzander wrote:
> On Mon, Feb 13, 2017 at 04:09:13PM -0200, Marcelo Tosatti wrote:
> > On Mon, Feb 13, 2017 at 03:42:04PM +0800, Eli Qiao wrote:
> > > > > L3data:0=0xff00;...
On Mon, Feb 13, 2017 at 04:09:13PM -0200, Marcelo Tosatti wrote:
On Mon, Feb 13, 2017 at 03:42:04PM +0800, Eli Qiao wrote:
> > L3data:0=0xff00;...
> > L3code:0=0xff00;...
> >
> > * Would you please help to test if the functions work.
>
> Setting up CDP machine.
>
> Unrelated:
>
> Found a bug:
>
68 matches
Mail list logo