[libvirt] [PATCH 0/6 v8] Unprivleged SG_IO support
by Osier Yang
As a result of RFC [1], this implements the unprivleged SG_IO
support.
1/6 and 2/6 are already acked.
v7 - v8:
* Change the XML tag name from "cdbfilter" to "sgio",
and to leave enough room for future values, the values
of "sgio" are "filtered" and "unfiltered" now.
v6 - v7:
* No restoring of unpriv_sgio per Daniel's thought.
* Use "major:minor" as the hash key per Jirka's suggestion.
Osier Yang (6):
util: Prepare helpers for unpriv_sgio setting
qemu: Add a hash table for the shared disks
docs: Add docs and rng schema for new XML tag sgio
conf: Parse and format the new XML
qemu: qemu: set unpriv_sgio when starting domain and attaching disk
qemu: Check if the shared disk's cdbfilter conflicts with others
docs/formatdomain.html.in | 14 ++-
docs/schemas/domaincommon.rng | 54 +++++---
src/conf/domain_conf.c | 55 ++++++--
src/conf/domain_conf.h | 10 ++
src/libvirt_private.syms | 4 +
src/qemu/qemu_conf.c | 86 ++++++++++++
src/qemu/qemu_conf.h | 12 ++
src/qemu/qemu_driver.c | 30 ++++
src/qemu/qemu_process.c | 99 ++++++++++++++
src/qemu/qemu_process.h | 4 +
src/util/util.c | 140 ++++++++++++++++++++
src/util/util.h | 13 ++
...qemuxml2argv-disk-scsi-lun-passthrough-sgio.xml | 32 +++++
tests/qemuxml2xmltest.c | 1 +
14 files changed, 522 insertions(+), 32 deletions(-)
create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-disk-scsi-lun-passthrough-sgio.xml
[1] https://www.redhat.com/archives/libvir-list/2012-November/msg00988.html
[2] https://www.redhat.com/archives/libvir-list/2012-December/msg00325.html
Regards,
Osier
12 years
[libvirt] [RFC] qcow2v3 features support
by Ján Tomko
Since 1.1 [1], QEMU supports qcow2 version 3 images. These are still
reffered to as qcow2 everywhere (qmp, qemu/qemu-img command line).
This format (besides some internal improvements) includes a bitfield of
compatible/incompatible/auto-clear features. [2]
The only feature so far is delayed refcount updating (since 1.2 [3]),
which is compatible with 1.1 if the file has been closed properly, if
not it has the dirty flag set.
For qcow2v3 images, -o compat=1.1 has to be passed to qemu-img.
(compat=0.10 means qcow2 which is the default at the moment)
For delayed refcount updates it's -o lazy_refcounts=on.
Using these options would require parsing 'qemu-img create -f qcow2 -o ?
/dev/null' output, or assuming they are supported.
Do we need a new format type (qcow2v3) that translates to qcow2 for
qemu, or would XML like this be enough?
<target>
<path>/var/lib/libvirt/images/image</path>
<format type='qcow2'/>
<features version='1.1'>
<lazy_refcounts compatible="no"/>
</features>
...
</target>
Jan
[1] http://git.qemu.org/?p=qemu.git;a=commitdiff;h=6744cbab8
[2] http://lists.gnu.org/archive/html/qemu-devel/2012-04/msg03077.html
[3] http://git.qemu.org/?p=qemu.git;a=commitdiff;h=bfe8043e9
12 years
[libvirt] libvirt domcument question
by zhijun liu
hi,all
when I read the document at http://libvirt.org/formatdomain.html about the
domain's xml format ,"disk" element has a optional subelment is
"readonly",when
I read the paragraph:readonlyIf present, this indicates the device cannot
be modified by the guest. For now, this is the default for disks with
attribute type='cdrom'.
it's say type='cdrom',I'm not sure that is correct or not,I think this is
a slips of pen and a possible writting maybe device='cdrom'.forgive me if I
misundestanded it. also I'm not sure should send this question to
libvirt-list or libvirt-user or others?liuzhijun
12 years
[libvirt] [PATCH] docs: Fix documentation for readonly element
by Jiri Denemark
---
Pushed as trivial.
docs/formatdomain.html.in | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in
index 8d9ab9e..94df6f8 100644
--- a/docs/formatdomain.html.in
+++ b/docs/formatdomain.html.in
@@ -1631,7 +1631,7 @@
<dt><code>readonly</code></dt>
<dd>If present, this indicates the device cannot be modified by
the guest. For now, this is the default for disks with
- attribute <code>type='cdrom'</code>.
+ attribute <code>device='cdrom'</code>.
</dd>
<dt><code>shareable</code></dt>
<dd>If present, this indicates the device is expected to be shared
--
1.8.0.2
12 years
[libvirt] [PATCH] Fix parsing of vendor_id
by Ken ICHIKAWA
I have a problem about parsing vendor_id of domain XML.
When define vendor_id attribute,
why definition of fallback attribute is needed?
I explain below for example.
(I used virsh edit.)
Current domain xml state is like below,
<domain>
<cpu>
</cpu>
</domain>
And I redefine domain xml like below,
<domain>
<cpu>
<model vendor_id='aaaabbbbcccc'>core2duo</model>
</cpu>
</domain>
Then, do dumpxml,
vendor_id is not reflected like below.
<domain>
<cpu mode='custom' match='exact'>
<model fallback='allow'>core2duo</model>
</cpu>
</domain>
I think this is not right behavior. It should be defined like
below.
<domain>
<cpu mode='custom' match='exact'>
<model fallback='allow' vendor_id='aaaabbbbcccc'>core2duo</model>
</cpu>
</domain>
And if I define fallback attribute and vendor_id attribute at the
same time, or define vendor_id attribute after fallback attribute
is defined, vendor_id attribute is reflected normally.
Is it a bug? or is there some reason?
I read past mailing list's thread about the patch adding vendor_id
but I could not find the reason.
https://www.redhat.com/archives/libvir-list/2012-June/thread.html#00917
It seems that patch v1 doesn't need definition of fallback attribute
but v2 needs it.
If it's a bug, please consider to apply this patch.
This patch fixes a problem that vendor_id attribute can not be
defined when fallback attribute is not defined.
Signed-off-by: Ken ICHIKAWA <ichikawa.ken(a)jp.fujitsu.com>
---
src/conf/cpu_conf.c | 42 +++++++++++++++++++++---------------------
1 file changed, 21 insertions(+), 21 deletions(-)
diff --git a/src/conf/cpu_conf.c b/src/conf/cpu_conf.c
index 8cb54a3..6157ed7 100644
--- a/src/conf/cpu_conf.c
+++ b/src/conf/cpu_conf.c
@@ -300,32 +300,32 @@ virCPUDefParseXML(const xmlNodePtr node,
goto error;
}
}
+ }
- if (virXPathBoolean("boolean(./model[1]/@vendor_id)", ctxt)) {
- char *vendor_id;
-
- vendor_id = virXPathString("string(./model[1]/@vendor_id)",
- ctxt);
- if (!vendor_id ||
- strlen(vendor_id) != VIR_CPU_VENDOR_ID_LENGTH) {
- virReportError(VIR_ERR_XML_ERROR,
- _("vendor_id must be exactly"
- " %d characters long"),
- VIR_CPU_VENDOR_ID_LENGTH);
+ if (virXPathBoolean("boolean(./model[1]/@vendor_id)", ctxt)) {
+ char *vendor_id;
+
+ vendor_id = virXPathString("string(./model[1]/@vendor_id)",
+ ctxt);
+ if (!vendor_id ||
+ strlen(vendor_id) != VIR_CPU_VENDOR_ID_LENGTH) {
+ virReportError(VIR_ERR_XML_ERROR,
+ _("vendor_id must be exactly"
+ " %d characters long"),
+ VIR_CPU_VENDOR_ID_LENGTH);
+ VIR_FREE(vendor_id);
+ goto error;
+ }
+ /* ensure that the string can be passed to qemu*/
+ for (i = 0; i < strlen(vendor_id); i++) {
+ if (vendor_id[i]==',') {
+ virReportError(VIR_ERR_XML_ERROR, "%s",
+ _("vendor id is invalid"));
VIR_FREE(vendor_id);
goto error;
}
- /* ensure that the string can be passed to qemu*/
- for (i = 0; i < strlen(vendor_id); i++) {
- if (vendor_id[i]==',') {
- virReportError(VIR_ERR_XML_ERROR, "%s",
- _("vendor id is invalid"));
- VIR_FREE(vendor_id);
- goto error;
- }
- }
- def->vendor_id = vendor_id;
}
+ def->vendor_id = vendor_id;
}
}
--
1.7.11.7
12 years
[libvirt] [PATCH] hostusb: Move USB_DEVFS define to hostusb.h to fix the build
by Cole Robinson
---
Pushed since it's a trivial build breaker
src/util/hostusb.c | 1 -
src/util/hostusb.h | 2 ++
2 files changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/util/hostusb.c b/src/util/hostusb.c
index 24f925b..2eb80a0 100644
--- a/src/util/hostusb.c
+++ b/src/util/hostusb.c
@@ -38,7 +38,6 @@
#include "virterror_internal.h"
#define USB_SYSFS "/sys/bus/usb"
-#define USB_DEVFS "/dev/bus/usb/"
#define USB_ID_LEN 10 /* "1234 5678" */
#define USB_ADDR_LEN 8 /* "123:456" */
diff --git a/src/util/hostusb.h b/src/util/hostusb.h
index aee1526..ad2476d 100644
--- a/src/util/hostusb.h
+++ b/src/util/hostusb.h
@@ -25,6 +25,8 @@
# include "internal.h"
+# define USB_DEVFS "/dev/bus/usb/"
+
typedef struct _usbDevice usbDevice;
typedef struct _usbDeviceList usbDeviceList;
--
1.8.0.2
12 years
[libvirt] libvirt-1.0.1 and F18 problem
by Gene Czarcinski
I am sure this not just a 1.0.1 problem and there appears to be a
bugzilla report on it:
https://bugzilla.redhat.com/show_bug.cgi?id=869625
Given a F17 host and a F18 host, there is significantly different
behavior when running a guest (in my case a F18 guest). On F18, I
cannot ssh or scp to the host let alone any other system. Also,
although DHCPv6 is working and I am getting an IPv6 DHCP address
assigned. Also, I can use firefox to access an Internet website.
Oh, and host returns no response.
I am not sure what is wrong but I hope this is easy to fix.
Gene
12 years
[libvirt] [PATCH] network: don't require private addresses if dnsmasq uses SO_BINDTODEVICE
by Laine Stump
This is yet another refinement to the fix for CVE-2012-3411:
https://bugzilla.redhat.com/show_bug.cgi?id=833033
It turns out that it would be very intrusive to correctly backport the
entire --bind-dynamic option to older dnsmasq versions
(e.g. dnsmasq-2.48 that is used on RHEL6.x and CentOS 6.x), but very
simple to patch those versions to just use SO_BINDTODEVICE on all
their listening sockets (SO_BINDTODEVICE also has the desired effect
of permitting only traffic that was received on the interface(s) where
dnsmasq was set to listen.)
This patch modifies the dnsmasq capabilities detection to detect the
string:
--bind-interfaces with SO_BINDTODEVICE
in the output of "dnsmasq --version", and in that case realize that
using the old --bind-interfaces option is just as safe as
--bind-dynamic (and therefore *not* forbid creation of networks that
use public IP address ranges).
If -bind-dynamic is available, it is still preferred over
--bind-interfaces.
Note that this patch does no harm in upstream, or in any distro's
downstream if it happens to end up there, but builds for distros that
have a new enough dnsmasq to support --bind-dynamic do *NOT* need to
specifically backport this patch; it's only required for distro
releases that have dnsmasq too old to have --bind-dynamic (and those
distros will need to add the SO_BINDTODEVICE patch to dnsmasq,
*including the extra string in the --version output*, as well.
---
src/network/bridge_driver.c | 6 ++++--
src/util/dnsmasq.c | 15 ++++++++++++---
src/util/dnsmasq.h | 1 +
3 files changed, 17 insertions(+), 5 deletions(-)
diff --git a/src/network/bridge_driver.c b/src/network/bridge_driver.c
index c62e729..fdd9382 100644
--- a/src/network/bridge_driver.c
+++ b/src/network/bridge_driver.c
@@ -725,13 +725,15 @@ networkDnsmasqConfContents(virNetworkObjPtr network,
* dnsmasq doesn't have bind-dynamic, only allow listening on
* private/local IP addresses (see RFC1918/RFC3484/RFC4193)
*/
- if (!virSocketAddrIsPrivate(&tmpipdef->address)) {
+ if (!dnsmasqCapsGet(caps, DNSMASQ_CAPS_BINDTODEVICE) &&
+ !virSocketAddrIsPrivate(&tmpipdef->address)) {
unsigned long version = dnsmasqCapsGetVersion(caps);
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Publicly routable address %s is prohibited. "
"The version of dnsmasq on this host (%d.%d) doesn't "
- "support the bind-dynamic option, which is required "
+ "support the bind-dynamic option or use SO_BINDTODEVICE "
+ "on listening sockets, one of which is required "
"for safe operation on a publicly routable subnet "
"(see CVE-2012-3411). You must either upgrade dnsmasq, "
"or use a private/local subnet range for this network "
diff --git a/src/util/dnsmasq.c b/src/util/dnsmasq.c
index e8eab1e..ba51447 100644
--- a/src/util/dnsmasq.c
+++ b/src/util/dnsmasq.c
@@ -664,10 +664,19 @@ dnsmasqCapsSetFromBuffer(dnsmasqCapsPtr caps, const char *buf)
if (strstr(buf, "--bind-dynamic"))
dnsmasqCapsSet(caps, DNSMASQ_CAPS_BIND_DYNAMIC);
- VIR_INFO("dnsmasq version is %d.%d, --bind-dynamic is %s",
+ /* if this string is a part of the --version output, dnsmasq
+ * has been patched to use SO_BINDTODEVICE when listening,
+ * so that it will only accept requests that arrived on the
+ * listening interface(s)
+ */
+ if (strstr(buf, "--bind-interfaces with SO_BINDTODEVICE"))
+ dnsmasqCapsSet(caps, DNSMASQ_CAPS_BINDTODEVICE);
+
+ VIR_INFO("dnsmasq version is %d.%d, --bind-dynamic is %spresent, "
+ "SO_BINDTODEVICE is %sin use",
(int)caps->version / 1000000, (int)(caps->version % 1000000) / 1000,
- dnsmasqCapsGet(caps, DNSMASQ_CAPS_BIND_DYNAMIC)
- ? "present" : "NOT present");
+ dnsmasqCapsGet(caps, DNSMASQ_CAPS_BIND_DYNAMIC) ? "" : "NOT ",
+ dnsmasqCapsGet(caps, DNSMASQ_CAPS_BIND_DYNAMIC) ? "" : "NOT ");
return 0;
fail:
diff --git a/src/util/dnsmasq.h b/src/util/dnsmasq.h
index 7a39232..b83bc96 100644
--- a/src/util/dnsmasq.h
+++ b/src/util/dnsmasq.h
@@ -68,6 +68,7 @@ typedef struct
typedef enum {
DNSMASQ_CAPS_BIND_DYNAMIC = 0, /* support for --bind-dynamic */
+ DNSMASQ_CAPS_BINDTODEVICE = 1, /* uses SO_BINDTODEVICE for --bind-interfaces */
DNSMASQ_CAPS_LAST, /* this must always be the last item */
} dnsmasqCapsFlags;
--
1.7.11.7
12 years
[libvirt] [PATCH] build: Fix AUTHORS generation
by Jiri Denemark
Using s/#authorslist#/$$out/ makes perl eat @domain part of all email
addresses from $out since it tries to interpret them as array variables.
I'm not sure if we can escape those in s/// but I know we can use print:
s/#authorslist#// and print '$$out'
to tell perl not to even inside $out.
This patch also fixes gen-AUTHORS so that it works in VPATH.
---
Makefile.am | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/Makefile.am b/Makefile.am
index 9556b42..05301cd 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -95,9 +95,9 @@ gen-ChangeLog:
.PHONY: gen-AUTHORS
gen-AUTHORS:
- $(AM_V_GEN)if test -d .git; then \
- out="`git log --pretty=format:'%aN <%aE>' | sort -u`" && \
- cat $(srcdir)/AUTHORS.in | perl -p -e "s/#authorslist#/$$out/" > \
+ $(AM_V_GEN)if test -d $(srcdir)/.git; then \
+ out="`cd $(srcdir) && git log --pretty=format:'%aN <%aE>' | sort -u`" && \
+ cat $(srcdir)/AUTHORS.in | perl -p -e "s/#authorslist#// and print '$$out'" > \
$(distdir)/AUTHORS-tmp && \
mv -f $(distdir)/AUTHORS-tmp $(distdir)/AUTHORS ; \
fi
--
1.8.0.2
12 years
