- Devel - libvirt List Archives

[Libvir] [Q] gnulib comment
by Atsushi SAKAI 28 Feb '08

28 Feb '08

Hi, Jim I have two questions on gnulib in libvirt. 1)tupe is typo?(I attach a patch.) (I do not guess the correct words, it is jargon?) 2)many NUL appeared on comment. Is the normal comment on gnulib? I think it should be NULL. c.f. I do spell check for my diversion. Thanks Atsushi SAKAI

4 3

[Libvir] PATCH: Set mac address on TUN device to fix Xenner
by Daniel P. Berrange 28 Feb '08

28 Feb '08

The QEMU driver pre-allocates a TAP device and connects it to a bridge when using virtual networking. The FD associated with the tap device is then passed to QEMU. When running Xen guests, Xenner needs to turn this FD back into a TUN device name. It does this by querying the hardware address of each NIC and looking for the one which matches that it expects - ie its TUN device. This obviously requires that the TUN device be configured with a MAC address though. libvirt does not currently do this. This patch fixes this and thus makes Xenner work from the QEMU driver when using virtual networking With this patch applied, and the latest 'virt-install' from upstream and the latest 'xenner' RPM from Fedora 9, it is now possible to create Xen paravirt guests under KVM using virt-install. That said there's a few more pending patch to cleanup various hacks. bridge.c | 14 ++++++++++++++ bridge.h | 1 + qemu_conf.c | 1 + 3 files changed, 16 insertions(+) Dan. Index: bridge.c =================================================================== RCS file: /data/cvs/libvirt/src/bridge.c,v retrieving revision 1.8 diff -u -p -r1.8 bridge.c --- bridge.c 5 Feb 2008 19:27:37 -0000 1.8 +++ bridge.c 27 Feb 2008 21:32:44 -0000 @@ -42,6 +42,7 @@ #include <linux/sockios.h> /* SIOCBRADDBR etc. */ #include <linux/if_bridge.h> /* SYSFS_BRIDGE_ATTR */ #include <linux/if_tun.h> /* IFF_TUN, IFF_NO_PI */ +#include <net/if_arp.h> /* ARPHRD_ETHER */ #include "internal.h" @@ -312,6 +313,7 @@ brDeleteInterface(brControl *ctl ATTRIBU int brAddTap(brControl *ctl, const char *bridge, + unsigned char *macaddr, char *ifname, int maxlen, int *tapfd) @@ -355,6 +357,18 @@ brAddTap(brControl *ctl, } if (ioctl(fd, TUNSETIFF, &try) == 0) { + struct ifreq addr; + memset(&addr, 0, sizeof(addr)); + memcpy(addr.ifr_hwaddr.sa_data, macaddr, 6); + addr.ifr_hwaddr.sa_family = ARPHRD_ETHER; + + /* Device actually starts in 'UP' state, but it + * needs to be down to set the MAC addr + */ + if ((errno = brSetInterfaceUp(ctl, try.ifr_name, 0))) + goto error; + if (ioctl(fd, SIOCSIFHWADDR, &addr) != 0) + goto error; if ((errno = brAddInterface(ctl, bridge, try.ifr_name))) goto error; if ((errno = brSetInterfaceUp(ctl, try.ifr_name, 1))) Index: bridge.h =================================================================== RCS file: /data/cvs/libvirt/src/bridge.h,v retrieving revision 1.4 diff -u -p -r1.4 bridge.h --- bridge.h 29 Jan 2008 18:15:54 -0000 1.4 +++ bridge.h 27 Feb 2008 21:32:44 -0000 @@ -62,6 +62,7 @@ int brDeleteInterface (brContr int brAddTap (brControl *ctl, const char *bridge, + unsigned char *mac, char *ifname, int maxlen, int *tapfd); Index: qemu_conf.c =================================================================== RCS file: /data/cvs/libvirt/src/qemu_conf.c,v retrieving revision 1.38 diff -u -p -r1.38 qemu_conf.c --- qemu_conf.c 27 Feb 2008 04:35:08 -0000 1.38 +++ qemu_conf.c 27 Feb 2008 21:32:44 -0000 @@ -1530,6 +1530,7 @@ qemudNetworkIfaceConnect(virConnectPtr c } if ((err = brAddTap(driver->brctl, brname, + net->mac, ifname, BR_IFNAME_MAXLEN, &tapfd))) { qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, "Failed to add tap interface '%s' to bridge '%s' : %s", -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|

3 2

[Libvir] [PATCH] Use virErrorMsg in qemudReportError
by Cole Robinson 28 Feb '08

28 Feb '08

Currently the qemu driver doesn't use virErrorMsg when raising error messages. This causes some errors to appear as "libvir: QEMU error: " if a custom string wasn't raised, which isn't all that useful. The patch below fixes this. Thanks, Cole diff --git a/src/qemu_conf.c b/src/qemu_conf.c index e39c7bc..6aed88b 100644 --- a/src/qemu_conf.c +++ b/src/qemu_conf.c @@ -58,6 +58,7 @@ void qemudReportError(virConnectPtr conn, int code, const char *fmt, ...) { va_list args; char errorMessage[QEMUD_MAX_ERROR_LEN]; + const char *virerr; if (fmt) { va_start(args, fmt); @@ -66,8 +67,10 @@ void qemudReportError(virConnectPtr conn, } else { errorMessage[0] = '\0'; } + + virerr = __virErrorMsg(code, (errorMessage[0] ? errorMessage[0] : NULL)); __virRaiseError(conn, dom, net, VIR_FROM_QEMU, code, VIR_ERR_ERROR, - NULL, NULL, NULL, -1, -1, "%s", errorMessage); + virerr, errorMessage, NULL, -1, -1, virerr, errorMessage); } int qemudLoadDriverConfig(struct qemud_driver *driver,

3 3

[Libvir] [PATCH] make qemud/'s rpcgen rules work
by Jim Meyering 27 Feb '08

27 Feb '08

I applied Rich's patch and found that qemud/remote_protocol.c was not automatically rebuilt. Since the Makefile rule is an implicit one using a non-default suffix (.x), you have to tell make about the new suffix. In straight make, you'd have to add ".SUFFIXES: .x" but with automake, it is recommended to use the SUFFIXES variable: Make qemud/'s rpcgen rules work. * qemud/Makefile.am (SUFFIXES): Define to .x before implicit rule. Signed-off-by: Jim Meyering <meyering(a)redhat.com> --- qemud/Makefile.am | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/qemud/Makefile.am b/qemud/Makefile.am index 1e1f861..3094b18 100644 --- a/qemud/Makefile.am +++ b/qemud/Makefile.am @@ -14,6 +14,7 @@ EXTRA_DIST = libvirtd.init.in libvirtd.sysconf default-network.xml \ libvirtd.conf if RPCGEN +SUFFIXES = .x .x.c: rm -f $@ rpcgen -c -o $@ $< -- 1.5.4.3.231.gc0a6

2 2

[Libvir] [PATCH] (for discussion) DHCP host mappings using 3 arrays API
by Richard W.M. Jones 27 Feb '08

27 Feb '08

This patch is an evolution of the previous patch for implementing DHCP host mappings. The virNetworkListDHCPHostMappings calls has been changed to take three arrays of strings: int virNetworkListDHCPHostMappings (virNetworkPtr network, char **const hwaddrs, char **const ipaddrs, char **const hostnames, int maxmappings); Also I've fixed a number of bugs throughout, particularly in the way that the hosts file was being parsed. This patch is for discussion only. There is a segfault on the client side of (perhaps ironically) the virNetworkListDHCPHostMappings call. Been looking for about an hour but I can't see it yet ... All the other calls appear to work OK. Rich. -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones virt-p2v converts physical machines to virtual machines. Boot with a live CD or over the network (PXE) and turn machines into Xen guests. http://et.redhat.com/~rjones/virt-p2v

4 8

[Libvir] [PATCH] Rewrite test-coverage rules to accommodate multiple .o files per .c.
by Jim Meyering 27 Feb '08

27 Feb '08

This makes the "make -C src cov" rule work with the addition of the libvirt_test.la convenience library. The trouble is that since there are now two .o files per .c (one for the "real" libvirt library, and another for the convenience library, libvirt_test.la, with no access restrictions, for testing), when you generate test-coverage stats, you have stats for each .c file in two different places, one for each .o file. This solves the problem by running gcov on each pair of .o files. A minor twist: when profiling with --disable-shared, the .o files are in ".", rather than in ".libs/", so configure.in now records the dirname in a new AC_SUBST'd variable, LV_LIBTOOL_OBJDIR, and src/Makefile.am uses that. Thus, profiling works with shared as well as static libraries. Rewrite test-coverage rules to accommodate multiple .o files per .c. * src/Makefile.am (cov): Rewrite rule to merge gcov results corresponding to two .o files: the libvirt_la- one, and the libvirt_test_la- one. (tst): Remove unused rule. * configure.in (LV_LIBTOOL_OBJDIR): Define and AC_SUBST. Signed-off-by: Jim Meyering <meyering(a)redhat.com> --- configure.in | 7 +++++++ src/Makefile.am | 24 ++++++++++++------------ 2 files changed, 19 insertions(+), 12 deletions(-) diff --git a/configure.in b/configure.in index c2030c7..6367990 100644 --- a/configure.in +++ b/configure.in @@ -913,6 +913,13 @@ AC_SUBST(MINGW_EXTRA_LDFLAGS) AC_SYS_LARGEFILE +# Set LV_LIBTOOL_OBJDIR to "." or $lt_cv_objdir, depending on whether +# we're building shared libraries. This is the name of the directory +# in which .o files will be created. +test "$enable_shared" = no && lt_cv_objdir=. +LV_LIBTOOL_OBJDIR=${lt_cv_objdir-.} +AC_SUBST(LV_LIBTOOL_OBJDIR) + # very annoying rm -f COPYING cp COPYING.LIB COPYING diff --git a/src/Makefile.am b/src/Makefile.am index c15de77..f8f2c4b 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -137,20 +137,20 @@ else EXTRA_DIST += parthelper.c endif -# -# target to ease building test programs -# -tst: tst.c - $(CC) $(CFLAGS) $(INCLUDES) -I../include -o tst tst.c .libs/libvirt.a $(LIBXML_LIBS) $(VIRSH_LIBS) $(GNUTLS_LIBS) $(LIBS) - -COVERAGE_FILES = $(CLIENT_SOURCES:%.c=libvirt_la-%.cov) - -cov: clean-cov $(COVERAGE_FILES) +cov: clean-cov + for i in $(CLIENT_SOURCES); do \ + case $$i in *.c) ;; *) continue;; esac; \ + b=$$(basename $$i .c); \ + o_files=; \ + for i in '' _test; do \ + o="$(LV_LIBTOOL_OBJDIR)/libvirt$${i}_la-$$b.o"; \ + test -f "$$o" \ + && o_files="$$o_files $$o"; \ + done; \ + gcov -b -f $$o_files > $$b.cov; \ + done clean-cov: rm -f *.cov *.gcov -%.cov: .libs/%.o - gcov -b -f -o .libs $< > $@ - CLEANFILES = *.cov *.gcov .libs/*.gcda .libs/*.gcno *.gcno *.gcda -- 1.5.4.3.231.gc0a6

1 0

[Libvir] [PATCH] Avoid "make distcheck" failure.
by Jim Meyering 27 Feb '08

27 Feb '08

Here's another trivial one: util-lib.h wasn't included in the tarball created by "make dist", so "make distcheck" would fail. Avoid "make distcheck" failure. * src/Makefile.am (virsh_SOURCES): Add util-lib.h. Signed-off-by: Jim Meyering <meyering(a)redhat.com> --- src/Makefile.am | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/Makefile.am b/src/Makefile.am index 02be9db..9c4810a 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -101,7 +101,7 @@ libvirt_la_CFLAGS = $(COVERAGE_CFLAGS) -DIN_LIBVIRT bin_PROGRAMS = virsh -virsh_SOURCES = virsh.c console.c console.h util-lib.c +virsh_SOURCES = virsh.c console.c console.h util-lib.c util-lib.h virsh_LDFLAGS = $(WARN_CFLAGS) $(COVERAGE_LDFLAGS) virsh_DEPENDENCIES = $(DEPS) virsh_LDADD = $(LDADDS) $(VIRSH_LIBS) -- 1.5.4.3.231.gc0a6

2 1

[Libvir] [PATCH] Fix a "make syntax-check" failure.
by Jim Meyering 27 Feb '08

27 Feb '08

"make syntax-check" reported this: src/capabilities.c Makefile.maint: the above files do not include <config.h> make: *** [sc_require_config_h] Error 1 The fix is so trivial it's barely worth posting, but just in case... And besides, it's good policy. Fix a "make syntax-check" failure. * src/capabilities.c: Include <config.h> from the .c file, * src/capabilities.h: ... not from the .h file. --- src/capabilities.c | 4 +++- src/capabilities.h | 3 --- 2 files changed, 3 insertions(+), 4 deletions(-) diff --git a/src/capabilities.c b/src/capabilities.c index 94f7d17..263935c 100644 --- a/src/capabilities.c +++ b/src/capabilities.c @@ -21,6 +21,8 @@ * Author: Daniel P. Berrange <berrange(a)redhat.com> */ +#include <config.h> + #include "capabilities.h" #include "buf.h" @@ -30,7 +32,7 @@ * @arch: host machine architecture * @offlineMigrate: non-zero if offline migration is available * @liveMigrate: non-zero if live migration is available - * + * * Allocate a new capabilities object */ virCapsPtr diff --git a/src/capabilities.h b/src/capabilities.h index 97bd06f..b379e51 100644 --- a/src/capabilities.h +++ b/src/capabilities.h @@ -24,9 +24,6 @@ #ifndef __VIR_CAPABILITIES_H #define __VIR_CAPABILITIES_H -#include <config.h> - - typedef struct _virCapsGuestFeature virCapsGuestFeature; typedef virCapsGuestFeature *virCapsGuestFeaturePtr; struct _virCapsGuestFeature { -- 1.5.4.3.231.gc0a6

1 0

[Libvir] [PATCH] fix typo
by Atsushi SAKAI 27 Feb '08

27 Feb '08

2 1

[Libvir] [PATCH 0/2] Re-organize generation / handling of capabilities data
by Daniel P. Berrange 27 Feb '08

27 Feb '08

Earlier today I started the 'simple' task of extending the QEMU driver to support Xenner for running Xen paravirt guests under KVM. This was at first glance easy because it basically looks like QEMU - in fact all we needed was to extend the capabilities XML for the QEMU driver to include data about Xen guests if Xenner is available. Unfortunately this was easier said than done... - The code generating the XML in qemu_driver.c was a mix of string formatting for the XML, and functional logic for determining runtime capabilities (kvm, kqemu, etc). - More runtime logic was in the qemu_conf.c file and stuff that should be runtime logic was static. I attempted to fix up the QEMU code for capabilities but it quickly turned into a horrible mess of spaghetti. Looking at the Xen driver, the situation was pretty much the same, but with the capabilities XML generation spread out over 3 files (xml.c, xen_internal.c and xend_internal.c). So I decided that we needed to have a internal API & structure to allow drivers to formally register & query their capabilities , and also add a central method for serializing it to XML. The following 2 patches implement this idea, and porting the Xen, QEMU and Test drivers to use it. This removes a lot of code duplication in XML formatting and makes the resulting code easier to follow, and should make it much quicker to adding capabilities info to new drivers too. Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|

3 10