[libvirt] [PATCH 00/11] Integrate usage of glib into libvirt
by Daniel P. Berrangé
This is a followup to a previous patch series:
https://www.redhat.com/archives/libvir-list/2019-August/msg01374.html
The first abort-on-oom parts of that series merged already.
As well as fixing the issues mentioned last time, the glib parts now do
a little more:
- Demonstrate conversion of virObject to GObject
- Convert to gbase64 APIs
- Start to convert getopt to GOptionContext
I have explicitly now confirmed we can freely mix g_malloc/malloc and
g_free/free, given our min glib version.
my intention with the glib code will be to focus on converting bits of
code that allow us to eliminate gnulib modules.
About 50% of the gnulib stuff is related to the Windows portability
for sockets() and poll(). We can address this by integrate the
event loop with GMainLoop, and using GIO for its GSocket APIs.
The other gnulib stuff is a random bag of APIs. Some may be replaced
by glib APIs, for others we can pull the gnulib fix straight into
libvirt, for others the portability problems might be obsolete.
Daniel P. Berrangé (11):
build: probe for glib-2 library in configure
build: link to glib, gobject, gio libraries
remote: don't pull anonymous enums into rpc protocol structs
util: use glib memory allocation functions
util: use glib string allocation/formatting functions
util: use glib base64 encoding/decoding APIs
util: convert virIdentity class to use GObject
src: convert over to use GRegex for regular exprssions
virsh: convert command line parsing to use GOptionContext
virt-admin: convert command line parsing to use GOptionContext
virt-login-shell: convert command line parsing to use GOptionContext
bootstrap.conf | 5 -
configure.ac | 2 +
libvirt.spec.in | 1 +
m4/virt-glib.m4 | 36 ++++
mingw-libvirt.spec.in | 2 +
src/Makefile.am | 5 +-
src/access/viraccessdriverpolkit.c | 21 +-
src/admin/admin_server.c | 3 +-
src/admin_protocol-structs | 9 -
src/conf/domain_event.c | 25 +--
src/conf/virsecretobj.c | 38 +---
src/internal.h | 1 +
src/libvirt_private.syms | 1 -
src/libxl/libxl_capabilities.c | 44 ++--
src/libxl/libxl_conf.c | 3 +-
src/lxc/Makefile.inc.am | 2 +
src/qemu/qemu_agent.c | 9 +-
src/qemu/qemu_command.c | 5 +-
src/qemu/qemu_domain.c | 8 +-
src/qemu/qemu_process.c | 4 +-
src/remote/Makefile.inc.am | 1 +
src/remote/remote_daemon.c | 3 +-
src/remote/remote_daemon_dispatch.c | 35 ++--
src/remote_protocol-structs | 9 -
src/rpc/virnetserverclient.c | 57 +++---
src/rpc/virnetserverprogram.c | 13 +-
src/secret/secret_driver.c | 1 -
src/storage/storage_backend_rbd.c | 4 +-
src/util/Makefile.inc.am | 1 +
src/util/viralloc.c | 29 +--
src/util/viridentity.c | 87 ++++----
src/util/viridentity.h | 7 +-
src/util/virstring.c | 40 +---
src/util/virstring.h | 2 -
tests/Makefile.am | 3 +-
tests/viridentitytest.c | 45 ++---
tests/virnetserverclienttest.c | 3 +-
tools/Makefile.am | 1 +
tools/virsh-secret.c | 17 +-
tools/virsh.c | 303 +++++++++++++---------------
tools/virt-admin.c | 207 +++++++++----------
tools/virt-login-shell-helper.c | 66 ++----
42 files changed, 495 insertions(+), 663 deletions(-)
create mode 100644 m4/virt-glib.m4
--
2.21.0
5 years, 1 month
[libvirt] [RFC] cpu_map: Remove pconfig from Icelake-Server CPU model
by Eduardo Habkost
The pconfig feature never worked, and adding "pconfig=off" to the
QEMU command-line triggers a regression in QEMU 3.1.1 and 4.0.0.
Signed-off-by: Eduardo Habkost <ehabkost(a)redhat.com>
---
I'm sending this as an RFC because I couldn't test it properly,
and because I don't know what are the consequences of changing
cpu_map between libvirt versions.
---
src/cpu_map/x86_Icelake-Server.xml | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/src/cpu_map/x86_Icelake-Server.xml b/src/cpu_map/x86_Icelake-Server.xml
index fb15977a59..188a781282 100644
--- a/src/cpu_map/x86_Icelake-Server.xml
+++ b/src/cpu_map/x86_Icelake-Server.xml
@@ -56,7 +56,9 @@
<feature name='pat'/>
<feature name='pcid'/>
<feature name='pclmuldq'/>
- <feature name='pconfig'/>
+ <!-- 'pconfig' was added by accident in QEMU 3.1.0 but never worked.
+ It was removed in QEMU 3.1.1 and 4.0.0. See QEMU commits
+ 76e5a4d58357 and 712f807e1965 for details -->
<feature name='pdpe1gb'/>
<feature name='pge'/>
<feature name='pku'/>
--
2.21.0
5 years, 1 month
[libvirt] [PATCH v4 00/12] remove repetition of URI path validation
by Daniel Henrique Barboza
This is a code repetition that I crossed a few times, then
I noticed that Cole Robinson suggested a solution for it
in the wiki. Here it is.
changes from v3:
- patch 8: fix the exception logic, move the code formatting
to patch 1
- patch 9: use lowcase 'qemu'
- patch 12: (optional) test case I created to aid in patch 8
logic
changes from v2:
- use a boolean to determine 'QEMU' and 'vbox' case to avoid block
repetition (patch 8)
- avoid 80+ chars lines in all patches
changes from v1:
- handle QEMU and vbox cases separately inside the validation
function
v3: https://www.redhat.com/archives/libvir-list/2019-September/msg01122.html
v2: https://www.redhat.com/archives/libvir-list/2019-September/msg01007.html
v1: https://www.redhat.com/archives/libvir-list/2019-September/msg00983.html
Daniel Henrique Barboza (12):
src/driver.c: add virConnectValidateURIPath()
interface_backend_netcf.c: use virConnectValidateURIPath()
interface_backend_udev.c: use virConnectValidateURIPath()
bridge_driver.c: virConnectValidateURIPath()
node_device_driver.c: use virConnectValidateURIPath()
secret_driver.c: use virConnectValidateURIPath()
storage_driver.c: use virConnectValidateURIPath()
driver.c: change URI validation to handle QEMU and vbox case
qemu_driver.c: use virConnectValidateURIPath()
vbox_common.c: use virConnectValidateURIPath()
vbox_driver.c: use virConnectValidateURIPath()
tests: add a test for driver.c:virConnectValidateURIPath()
src/driver.c | 38 +++++
src/driver.h | 4 +
src/interface/interface_backend_netcf.c | 19 +--
src/interface/interface_backend_udev.c | 19 +--
src/libvirt_private.syms | 1 +
src/network/bridge_driver.c | 19 +--
src/node_device/node_device_driver.c | 19 +--
src/qemu/qemu_driver.c | 20 +--
src/secret/secret_driver.c | 19 +--
src/storage/storage_driver.c | 19 +--
src/vbox/vbox_common.c | 16 +-
src/vbox/vbox_driver.c | 16 +-
tests/Makefile.am | 7 +-
tests/virdriverconnvalidatetest.c | 186 ++++++++++++++++++++++++
14 files changed, 267 insertions(+), 135 deletions(-)
create mode 100644 tests/virdriverconnvalidatetest.c
--
2.21.0
5 years, 1 month
[libvirt] [PATCH 1/1] src/driver.c: remove duplicated code in virGetConnect* functions
by Daniel Henrique Barboza
All the 6 virGetConnect* functions in driver.c shares the
same code base. This patch creates a new static function
virGetConnectGeneric() that contains the common code to
be used with all other virGetConnect*.
Signed-off-by: Daniel Henrique Barboza <danielhb413(a)gmail.com>
---
CC'ing Cole Robinson since he reviewed similar patches
a few days ago.
src/driver.c | 100 +++++++++++++--------------------------------------
1 file changed, 25 insertions(+), 75 deletions(-)
diff --git a/src/driver.c b/src/driver.c
index ed2d943ddf..8a4bc8ff66 100644
--- a/src/driver.c
+++ b/src/driver.c
@@ -29,6 +29,7 @@
#include "virfile.h"
#include "virlog.h"
#include "virmodule.h"
+#include "virstring.h"
#include "virthread.h"
#include "configmake.h"
@@ -96,112 +97,61 @@ virConnectCacheOnceInit(void)
VIR_ONCE_GLOBAL_INIT(virConnectCache);
-virConnectPtr virGetConnectInterface(void)
+static virConnectPtr
+virGetConnectGeneric(virThreadLocal thread, const char *name)
{
virConnectPtr conn;
if (virConnectCacheInitialize() < 0)
return NULL;
- conn = virThreadLocalGet(&connectInterface);
+ conn = virThreadLocalGet(&thread);
+
if (conn) {
- VIR_DEBUG("Return cached interface connection %p", conn);
+ VIR_DEBUG("Return cached %s connection %p", name, conn);
virObjectRef(conn);
} else {
- conn = virConnectOpen(geteuid() == 0 ? "interface:///system" : "interface:///session");
- VIR_DEBUG("Opened new interface connection %p", conn);
+ VIR_AUTOFREE(char *) uri = NULL;
+ const char *uriPath = geteuid() == 0 ? "/system" : "/session";
+
+ if (virAsprintf(&uri, "%s//%s", name, uriPath) < 0)
+ return NULL;
+
+ conn = virConnectOpen(uri);
+ VIR_DEBUG("Opened new %s connection %p", name, conn);
}
return conn;
}
-virConnectPtr virGetConnectNetwork(void)
-{
- virConnectPtr conn;
- if (virConnectCacheInitialize() < 0)
- return NULL;
+virConnectPtr virGetConnectInterface(void)
+{
+ return virGetConnectGeneric(connectInterface, "interface");
+}
- conn = virThreadLocalGet(&connectNetwork);
- if (conn) {
- VIR_DEBUG("Return cached network connection %p", conn);
- virObjectRef(conn);
- } else {
- conn = virConnectOpen(geteuid() == 0 ? "network:///system" : "network:///session");
- VIR_DEBUG("Opened new network connection %p", conn);
- }
- return conn;
+virConnectPtr virGetConnectNetwork(void)
+{
+ return virGetConnectGeneric(connectNetwork, "network");
}
virConnectPtr virGetConnectNWFilter(void)
{
- virConnectPtr conn;
-
- if (virConnectCacheInitialize() < 0)
- return NULL;
-
- conn = virThreadLocalGet(&connectNWFilter);
- if (conn) {
- VIR_DEBUG("Return cached nwfilter connection %p", conn);
- virObjectRef(conn);
- } else {
- conn = virConnectOpen(geteuid() == 0 ? "nwfilter:///system" : "nwfilter:///session");
- VIR_DEBUG("Opened new nwfilter connection %p", conn);
- }
- return conn;
+ return virGetConnectGeneric(connectNWFilter, "nwfilter");
}
virConnectPtr virGetConnectNodeDev(void)
{
- virConnectPtr conn;
-
- if (virConnectCacheInitialize() < 0)
- return NULL;
-
- conn = virThreadLocalGet(&connectNodeDev);
- if (conn) {
- VIR_DEBUG("Return cached nodedev connection %p", conn);
- virObjectRef(conn);
- } else {
- conn = virConnectOpen(geteuid() == 0 ? "nodedev:///system" : "nodedev:///session");
- VIR_DEBUG("Opened new nodedev connection %p", conn);
- }
- return conn;
+ return virGetConnectGeneric(connectNodeDev, "nodedev");
}
virConnectPtr virGetConnectSecret(void)
{
- virConnectPtr conn;
-
- if (virConnectCacheInitialize() < 0)
- return NULL;
-
- conn = virThreadLocalGet(&connectSecret);
- if (conn) {
- VIR_DEBUG("Return cached secret connection %p", conn);
- virObjectRef(conn);
- } else {
- conn = virConnectOpen(geteuid() == 0 ? "secret:///system" : "secret:///session");
- VIR_DEBUG("Opened new secret connection %p", conn);
- }
- return conn;
+ return virGetConnectGeneric(connectSecret, "secret");
}
virConnectPtr virGetConnectStorage(void)
{
- virConnectPtr conn;
-
- if (virConnectCacheInitialize() < 0)
- return NULL;
-
- conn = virThreadLocalGet(&connectStorage);
- if (conn) {
- VIR_DEBUG("Return cached storage connection %p", conn);
- virObjectRef(conn);
- } else {
- conn = virConnectOpen(geteuid() == 0 ? "storage:///system" : "storage:///session");
- VIR_DEBUG("Opened new storage connection %p", conn);
- }
- return conn;
+ return virGetConnectGeneric(connectStorage, "storage");
}
--
2.21.0
5 years, 1 month
[libvirt] [PATCH v2 0/2] make attaching disk partition to VM illegal
by Pavel Mores
This version addresses comments and objections to the original submission.
Pavel Mores (2):
qemu: make attaching disk partition to VM illegal
remove a now redundant call to virDiskNameToIndex()
src/qemu/qemu_command.c | 6 -----
src/qemu/qemu_domain.c | 15 +++++++++++
.../disk-attaching-partition-nosupport.xml | 27 +++++++++++++++++++
tests/qemuxml2argvtest.c | 1 +
4 files changed, 43 insertions(+), 6 deletions(-)
create mode 100644 tests/qemuxml2argvdata/disk-attaching-partition-nosupport.xml
--
2.21.0
5 years, 1 month
[libvirt] [PATCH v2 0/6] rewrite virt-host-validate to be data driven, using Go & YAML
by Daniel P. Berrangé
This is a followup to a previous PoC patch I submitted a
month ago:
https://www.redhat.com/archives/libvir-list/2019-September/msg00036.html
The commit messages in the individual patches given quite a
bit of detail, so I'll keep this cover letter brief.
In my previous posting I was unhappy with the implications for
the RPM packaging, and was considering having this as a separate
source repo & RPM. On further investigation such an approach
would not in fact solve the RPM packaging problem, because we
would still not be using a pure go build toolchain, as we have
data files that need installing in the right place.
This forced me to actually address the RPM packaging problems
that Fedora had with Go when used from a build tool like make
or meson.
After alot of debugging I finally got a viable solution merged
into the Fedora go-rpm-macros package:
https://pagure.io/go-rpm-macros/c/67b4fbbbfce0986ac46cd1329bf85a18ea7a43d2
commit 67b4fbbbfce0986ac46cd1329bf85a18ea7a43d2
Author: Daniel P. Berrangé <berrange(a)redhat.com>
Date: Wed Sep 18 16:49:58 2019 +0100
macros: define a %gobuildflags macro
Using the %gobuild macro is fine for a project where the go
code is the only thing being built, and can be built directly
by invoking the Go toolchain from RPM.
In more complex cases though, the Go code is just a small part
of the project and the Go toolchain is invoked by a build
system such as make (possibly automake), or meson. In such a
case we need to be able to tell this build system what flags
to pass to the compiler.
The %gobuildflags macros services this purpose allowing a
RPM spec todo
GOBUILDFLAGS="%gobuildflags" %configure
or
%make GOBUILDFLAGS="%gobuildflags"
Ideally the %gobuild macro would in turn reference the
%gobuildflags macro, but that does not appear possible
given the semantics around quote expansion and escaping
across RPM and shell.
Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com>
As a result in this series, we're now fully integrated into the
RPM build, on Fedora at least. I've not checked what approach
RHEL takes for Go, whether it requires separate RPM for each
3rd party dep, or prefers bundling. Either way though, we can
deal with the problem now.
The other obvious change is that this is now a patch series,
to make it easier to review the code in managable chunks.
The really big difference though is that I replaced the use
of XML data files with YAML data files. This was done with
the aim of making the data more human friendly. XML is really
optimized for machines, not humans, so writing the data files
was not pretty. YAML is optimized for human readability, and
is actually even easier to consume in Go than the XML was,
so its a double win.
Finally, we also add new checks at the end for the various
CPU hardware side channel mitigations, and report whether
SMT/HT is unsafe or not (any Intel host is basically unsafe
before Icelake).
Daniel P. Berrangé (6):
build: introduce logic for using golang in libvirt
tools: introduce a data driven impl of virt-host-validate
tools: define YAML rules for virt-host-validate checks
tools: switch to build the new virt-host-validate impl
tools: delete the old virt-host-validate impl
tools: make virt-host-validate check CPU vulnerabilities
configure.ac | 1 +
libvirt.spec.in | 35 +-
m4/virt-golang.m4 | 46 ++
m4/virt-host-validate.m4 | 8 +-
po/POTFILES | 5 -
tools/Makefile.am | 76 +--
tools/host-validate/go.mod | 10 +
tools/host-validate/go.sum | 9 +
tools/host-validate/main.go | 98 +++
tools/host-validate/pkg/engine.go | 481 ++++++++++++++
tools/host-validate/pkg/facts.go | 585 ++++++++++++++++++
.../pkg/facts_test.go} | 36 +-
tools/host-validate/rules/builtin.yaml | 20 +
tools/host-validate/rules/cpu.yaml | 50 ++
tools/host-validate/rules/freebsd-kernel.yaml | 77 +++
tools/host-validate/rules/linux-acpi.yaml | 39 ++
tools/host-validate/rules/linux-cgroups.yaml | 470 ++++++++++++++
.../rules/linux-cpu-hardware-flaws.yaml | 165 +++++
tools/host-validate/rules/linux-cpu.yaml | 134 ++++
tools/host-validate/rules/linux-devices.yaml | 71 +++
tools/host-validate/rules/linux-iommu.yaml | 113 ++++
.../host-validate/rules/linux-namespaces.yaml | 119 ++++
tools/host-validate/rules/linux-pci.yaml | 10 +
tools/virt-host-validate-bhyve.c | 77 ---
tools/virt-host-validate-common.c | 419 -------------
tools/virt-host-validate-common.h | 85 ---
tools/virt-host-validate-lxc.c | 87 ---
tools/virt-host-validate-lxc.h | 24 -
tools/virt-host-validate-qemu.c | 116 ----
tools/virt-host-validate-qemu.h | 24 -
tools/virt-host-validate.c | 152 -----
tools/virt-host-validate.pod | 12 +-
32 files changed, 2609 insertions(+), 1045 deletions(-)
create mode 100644 m4/virt-golang.m4
create mode 100644 tools/host-validate/go.mod
create mode 100644 tools/host-validate/go.sum
create mode 100644 tools/host-validate/main.go
create mode 100644 tools/host-validate/pkg/engine.go
create mode 100644 tools/host-validate/pkg/facts.go
rename tools/{virt-host-validate-bhyve.h => host-validate/pkg/facts_test.go} (52%)
create mode 100644 tools/host-validate/rules/builtin.yaml
create mode 100644 tools/host-validate/rules/cpu.yaml
create mode 100644 tools/host-validate/rules/freebsd-kernel.yaml
create mode 100644 tools/host-validate/rules/linux-acpi.yaml
create mode 100644 tools/host-validate/rules/linux-cgroups.yaml
create mode 100644 tools/host-validate/rules/linux-cpu-hardware-flaws.yaml
create mode 100644 tools/host-validate/rules/linux-cpu.yaml
create mode 100644 tools/host-validate/rules/linux-devices.yaml
create mode 100644 tools/host-validate/rules/linux-iommu.yaml
create mode 100644 tools/host-validate/rules/linux-namespaces.yaml
create mode 100644 tools/host-validate/rules/linux-pci.yaml
delete mode 100644 tools/virt-host-validate-bhyve.c
delete mode 100644 tools/virt-host-validate-common.c
delete mode 100644 tools/virt-host-validate-common.h
delete mode 100644 tools/virt-host-validate-lxc.c
delete mode 100644 tools/virt-host-validate-lxc.h
delete mode 100644 tools/virt-host-validate-qemu.c
delete mode 100644 tools/virt-host-validate-qemu.h
delete mode 100644 tools/virt-host-validate.c
--
2.21.0
5 years, 1 month
[libvirt] [PATCH] docs: document that C & Python are the preferred languages
by Daniel P. Berrangé
Blacklist Perl and Shell code in favour of Python for
sake of readability and portability.
Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com>
---
docs/hacking.html.in | 30 ++++++++++++++++++++++++++++++
1 file changed, 30 insertions(+)
diff --git a/docs/hacking.html.in b/docs/hacking.html.in
index 90376968c3..8572eb65a6 100644
--- a/docs/hacking.html.in
+++ b/docs/hacking.html.in
@@ -343,6 +343,36 @@
Richard Jones' guide to working with open source projects</a>.
</p>
+ <h2><a id="lang">Language Usage</a></h2>
+
+ <p>
+ The libvirt repository makes use of a large number of programming
+ languages. There is a general desire to phase out some of the
+ existing languages used to reduce the knowledge burden on
+ developers, and facilitate introduction of new languages in
+ the future.
+ </p>
+
+ <p>
+ The preferred languages at this time are
+ </p>
+
+ <ul>
+ <li>C - for the main libvirt codebase. Dialect supported by
+ GCC/CLang only.</li>
+ <li>Python - for supporting build scripts / tools. Code must
+ run with both version 2.7 and 3.x at this time.</li>
+ </ul>
+
+ <p>
+ Languages that should not be used for any new contributions.
+ </p>
+
+ <ul>
+ <li>Perl - build scripts must be written in Python instead.</li>
+ <li>Shell - build scripts must be written in Python instead.</li>
+ </ul>
+
<h2><a id="tooling">Tooling</a></h2>
<p>
--
2.21.0
5 years, 1 month
[libvirt] [PATCH 0/2] qemu_process: use VIR_AUTO* all around
by Daniel Henrique Barboza
The usual AUTOFREE() and AUTOUNREF() changes that allows
for a bit of cleanup.
Daniel Henrique Barboza (2):
qemu_process: use VIR_AUTOFREE()
qemu_process: use VIR_AUTOUNREF()
src/qemu/qemu_process.c | 413 +++++++++++++++-------------------------
1 file changed, 149 insertions(+), 264 deletions(-)
--
2.21.0
5 years, 1 month
[libvirt] [PATCH] rpc: fix escaping of shell path for netcat binary
by Daniel P. Berrangé
Consider having a nc binary in the path with a space in its name,
for example '/tmp/fo o/nc'
This results in libvirt running SSH with the following arg value
"'if ''/tmp/fo o/nc'' -q 2>&1 | grep \"requires
an argument\" >/dev/null 2>&1; then ARG=-q0;
else ARG=;fi;''/tmp/fo o/nc'' $ARG -U
/var/run/libvirt/libvirt-sock'"
The use of the single quote escaping was introduced by
commit 6ac6238de33fc74e7545b245ae273d1bfd658808
Author: Guido Günther <agx(a)sigxcpu.org>
Date: Thu Oct 13 21:49:01 2011 +0200
Use virBufferEscapeShell in virNetSocketNewConnectSSH
to escape the netcat command since it's passed to the shell. Adjust
expected test case output accordingly.
While the intention of this change was good, the result is broken as it
is still underquoted.
On the SSH server side, SSH itself runs the command via the shell.
Our command is then invoking the shell again. Thus we see
$ virsh -c qemu+ssh://root@domokun/system?netcat=%2Ftmp%2Ffo%20o%2Fnc list
error: failed to connect to the hypervisor
error: End of file while reading data: sh: /tmp/fo: No such file or directory: Input/output error
With the second level of escaping added we can now successfully use a nc
binary with a space in the path.
The original test case added was misleading as it illustrated using a
binary path of 'nc -4' which is not a path, it is a command with a
separate argument, which is getting interpreted as a path.
Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com>
---
src/rpc/virnetclient.c | 8 ++++++++
src/rpc/virnetsocket.c | 9 +++++++++
tests/virnetsockettest.c | 6 +++---
3 files changed, 20 insertions(+), 3 deletions(-)
diff --git a/src/rpc/virnetclient.c b/src/rpc/virnetclient.c
index 64855fb8d6..53d8b219ea 100644
--- a/src/rpc/virnetclient.c
+++ b/src/rpc/virnetclient.c
@@ -490,6 +490,10 @@ virNetClientPtr virNetClientNewLibSSH2(const char *host,
DEFAULT_VALUE(knownHostsVerify, "normal");
virBufferEscapeShell(&buf, netcatPath);
+ if (!(nc = virBufferContentAndReset(&buf)))
+ goto no_memory;
+ virBufferEscapeShell(&buf, nc);
+ VIR_FREE(nc);
if (!(nc = virBufferContentAndReset(&buf)))
goto no_memory;
@@ -596,6 +600,10 @@ virNetClientPtr virNetClientNewLibssh(const char *host,
DEFAULT_VALUE(knownHostsVerify, "normal");
virBufferEscapeShell(&buf, netcatPath);
+ if (!(nc = virBufferContentAndReset(&buf)))
+ goto no_memory;
+ virBufferEscapeShell(&buf, nc);
+ VIR_FREE(nc);
if (!(nc = virBufferContentAndReset(&buf)))
goto no_memory;
diff --git a/src/rpc/virnetsocket.c b/src/rpc/virnetsocket.c
index ebd304707a..a469907779 100644
--- a/src/rpc/virnetsocket.c
+++ b/src/rpc/virnetsocket.c
@@ -903,6 +903,15 @@ int virNetSocketNewConnectSSH(const char *nodename,
return -1;
}
quoted = virBufferContentAndReset(&buf);
+
+ virBufferEscapeShell(&buf, quoted);
+ VIR_FREE(quoted);
+ if (virBufferCheckError(&buf) < 0) {
+ virCommandFree(cmd);
+ return -1;
+ }
+ quoted = virBufferContentAndReset(&buf);
+
/*
* This ugly thing is a shell script to detect availability of
* the -q option for 'nc': debian and suse based distros need this
diff --git a/tests/virnetsockettest.c b/tests/virnetsockettest.c
index bb8357f7cd..8cad351605 100644
--- a/tests/virnetsockettest.c
+++ b/tests/virnetsockettest.c
@@ -661,15 +661,15 @@ mymain(void)
struct testSSHData sshData7 = {
.nodename = "somehost",
- .netcat = "nc -4",
+ .netcat = "/tmp/fo o/nc",
.path = "/tmp/socket",
.expectOut = "-T -e none -- somehost sh -c '"
- "if ''nc -4'' -q 2>&1 | grep \"requires an argument\" >/dev/null 2>&1; then "
+ "if \'''\\''/tmp/fo o/nc'\\'''' -q 2>&1 | grep \"requires an argument\" >/dev/null 2>&1; then "
"ARG=-q0;"
"else "
"ARG=;"
"fi;"
- "''nc -4'' $ARG -U /tmp/socket'\n",
+ "'''\\''/tmp/fo o/nc'\\'''' $ARG -U /tmp/socket'\n",
};
if (virTestRun("SSH test 7", testSocketSSH, &sshData7) < 0)
ret = -1;
--
2.21.0
5 years, 1 month
[libvirt] [PATCH] admin: fix memory leak of typed parameters getting client info
by Daniel P. Berrangé
In the error code path, the temporary parameters are not freed.
Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com>
---
src/admin/admin_server.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/src/admin/admin_server.c b/src/admin/admin_server.c
index 248df3f795..0d6091937d 100644
--- a/src/admin/admin_server.c
+++ b/src/admin/admin_server.c
@@ -310,6 +310,8 @@ adminClientGetInfo(virNetServerClientPtr client,
ret = 0;
cleanup:
+ if (tmpparams)
+ virTypedParamsFree(tmpparams, *nparams);
virObjectUnref(identity);
VIR_FREE(sock_addr);
return ret;
--
2.21.0
5 years, 1 month
