- Devel - libvirt List Archives

[libvirt] [PATCH] Fix XML escaping problem in the XM driver
by Daniel Veillard 19 Jun '08

19 Jun '08

Basically the XM driver when generating the XML config forgets to escape things like filepaths allowing characters which are forbidden as-is in XML instances to pass in the resulting config files which then are unusable. The patch also adds a new test checking the escaping is done in a few critical places. Daniel -- Red Hat Virtualization group http://redhat.com/virtualization/ Daniel Veillard | virtualization library http://libvirt.org/ veillard(a)redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/

1 0

[libvirt] Test packages for various fixes
by Chris Lalancette 19 Jun '08

19 Jun '08

I've sent a build through koji that contains 0.4.3 + iSCSI fixes + VIR_ALLOC fixes + tapfd fixes for Fedora 9. It's available here, if you want to test: http://koji.fedoraproject.org/koji/taskinfo?taskID=670309 As usual, if there are any bugs that you find, please report them here. Chris Lalancette

1 0

[libvirt] [PATCH]: Fix allocation of tapfds when starting qemu
by Chris Lalancette 19 Jun '08

19 Jun '08

All, When doing the conversion to danpb's new memory API, a small bug was introduced into the qemudNetworkIfaceConnect() function. In particular, there is a call: if (VIR_ALLOC_N(vm->tapfds, vm->ntapfds+2) < 0) goto no_memory; However, the tapfds structure is used to track *all* of the tap fds, and is called once for each network that is being attached to the domain. VIR_ALLOC_N maps to calloc(). So the first network would work just fine, but if you had more than one network, subsequent calls to this function would blow away the stored fd's that were already there and fill them all in with zeros. This causes multiple problems, from the qemu domains not starting properly to improper cleanup on shutdown. The attached patch just changes the VIR_ALLOC_N() to a VIR_REALLOC_N(), and everything is happy again. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

4 7

[libvirt] [PATCH]: Fix VIR_ALLOC_N for 0 byte arrays
by Chris Lalancette 19 Jun '08

19 Jun '08

Hello, For 0.4.3, danpb's new memory management scheme went into libvirt. This is fine, except that is subtly alters the semantics of malloc(), calloc(), and realloc(). In particular, if you say: foo = malloc(0); glibc will happily return a non-NULL pointer to you. However, with the new memory management stuff, if you say: foo = VIR_ALLOC(0); you will actually get a NULL pointer back. Personally, I think this is a dangerous deviation from malloc() semantics that everyone is used to, and is indeed causing problems with the remote driver. The short of it is that the remote driver allocates memory on behalf of the remote side using VIR_ALLOC_N, and this call is returning NULL so that the NULL checks elsewhere in the code fire and return failure. The attached patch fixes this situation by removing the 0 checks from the memory allocation paths, and just lets them fall through to the normal malloc(), calloc(), or realloc() routines, restoring old semantics. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

4 3

[libvirt] How does libvirt access kvm infomation?
by Akio Takebe 19 Jun '08

19 Jun '08

Hi, AFAIK kvm has only an interface of qemu console when we use device hotplug, migration and so on. How does libvirt access the interface of qemu console? Best Regards, Akio Takebe

2 2

[libvirt] Cherokee+Libvirt+Avahi = Cool Libre Application
by Stefan de Konink 18 Jun '08

18 Jun '08

Hi, As I have told on the mailinglist before I'm working on a webbased implementation of the libvirt API. Using Avahi as messaging protocol, some extra tools were written: - Domumdns (Tool that puts every domain on the network (scanning interval of 10s connecting to the local hypervisor), refering back to the hostname of libvirtd) - Cherokee Avahi Handler, Cherokee Virt Handler (implements HTTP API) So basically: http://xen.bot.nu/avahi/ Shows the available vms on my network, every 10 seconds updated by domudmns. My proposal would be to add this functionality by default to libvirtd. 1) it is useful 2) it implements something of the todo list: events. http://xen.bot.nu/virt/klant1/gentoo/virDomainGetID http://xen.bot.nu/virt/klant1/gentoo/virDomainGetName http://xen.bot.nu/virt/klant1/gentoo/virDomainGetMaxMemory http://xen.bot.nu/virt/klant1/gentoo/virDomainGetMaxVcpus http://xen.bot.nu/virt/klant1/gentoo/virDomainGetOSType http://xen.bot.nu/virt/klant1/gentoo/virDomainGetUUID <-- that's odd http://xen.bot.nu/virt/klant1/gentoo/virDomainGetUUIDString http://xen.bot.nu/virt/klant1/gentoo/virDomainGetXMLDesc Are implemented now, the updates as attach/detach devices will be added soon, but that is not something I'm going to demo without some basic authentication ;) Stefan

1 0

[libvirt] [PATCH 5/5]: Rewrite findLuns function
by Chris Lalancette 17 Jun '08

17 Jun '08

This rather large patch rewrites the virStorageBackendISCSIFindLUNs() function to only rely on sysfs for finding LUNs, given a session number. Along the way, it also fixes the bug where we wouldn't find LUNs for older kernels (with the block:sda format), and also (possibly) fixes a race condition where we could try to find the LUN before udev has finished connecting it. I say it "possibly" fixes it because I haven't been able to hit it so far, but I definitely need more testing to try and confirm. Changes since last time: 1) Don't blindly ignore the 0'th LUN (thanks Stefan). Instead, look if the LUNs are Direct-Access devices (as defined by the kernel); if so, they are LUNs we can use. 2) Fix up whitespace damage. 3) Check the return value of the scsidev strdup as pointed out by Jim. 4) Change all ISCSIADM commands to be const char *const as pointed out by Jim. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

3 5

[libvirt] [PATCH 4/5]: Move the sendtarget command into the login command
by Chris Lalancette 17 Jun '08

17 Jun '08

A small bugfix; we only need to call the iscsiadm sendtarget command when we are first logging in; we don't need to do it for logout. Move the sendtarget command into the Login() function. Changes since last time: 1) Make const char *cmdsendtarget into const char *const cmdsendtarget based on feedback from Jim. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 2

[libvirt] [PATCH 3/5]: Fix up the call to iscsiadm --mode session
by Chris Lalancette 17 Jun '08

17 Jun '08

Older versions of iscsiadm didn't support the "-P 0" flag to the "iscsiadm --mode session" command. However, just running "iscsiadm --mode session" seems to work on all version of iscsiadm commands back to FC-6, so just use that. Changes since last time: None Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 2

[libvirt] [PATCH 2/5]: Cleanup carriage returns in util.c
by Chris Lalancette 17 Jun '08

17 Jun '08

In src/util.c, virLog is just a wrapper around fprintf(stderr). Make sure to put line breaks at the end of lines that use virLog() (noticed during testing). Changes since last time: None Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 2

[libvirt] [PATCH 1/5]: Return exitstatus from virStorageBackendRunProgRegex
by Chris Lalancette 17 Jun '08

17 Jun '08

This patch changes things around so that virStorageBackendRunProgRegex() does *not* virStorageReportError() if the fork()/exec() process it spawned returned a != 0 exit code. Rather, it returns the exitcode in this case, and it is up to the higher level to determine whether this is a fatal error or not. The use case for this change is in the iSCSI stuff; older versions of iscsiadm tools would return a failure when getting the session number, despite the command succeeding. Changes since last time: 1) Fix up whitespace damage Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 2

[libvirt] [PATCH 0/5]: Rework iSCSI lun handling
by Chris Lalancette 17 Jun '08

17 Jun '08

These patches are now better tested, so I'm sending them for review again. To recap, these patches: 1) Do minor cleanup of the existing code 2) Fix a bug where we aren't properly finding the LUNs on pre 2.6.24 kernels 3) Fix a bug seen in oVirt where we race between libvirtd scanning the sysfs path and udev actually making the sysfs path in the first place 4) Further lessen our dependence on direct calls to iscsiadm, in case they change the output in the future. Changes since last time include: 1) Fixing the 0th LUN problem as pointed out by Stefan de Konink. We now use a sysfs file to determine whether this is a valid LUN for us to have. 2) Fix all my whitespace damage 3) Small cleanups based on feedback from the list. Also, I'm putting a build through koji, available here: http://koji.fedoraproject.org/koji/taskinfo?taskID=663583 Chris Lalancette

4 6

[libvirt] [PATCH] Pass syntax check for block devices to XenD.
by Hiroyuki Kaguchi 17 Jun '08

17 Jun '08

Hi, libvirt doesn't pass disk information that the device type isn't specified to XenD and doesn't show the error message either. However, the error message should be displayed, and XenD should show it. On this account, I change libvirt that libvirt passes the disk information that the device type isn't specified to XenD and XenD shows error message. The related patch: After Xen Cset17617, XenD check the syntax of the device type of the disk information. Thanks, Signed-off-by: Hiroyuki Kaguchi <fj7025cf(a)aa.jp.fujitsu.com> Index: src/xm_internal.c =================================================================== RCS file: /data/cvs/libvirt/src/xm_internal.c,v retrieving revision 1.79 diff -u -p -r1.79 xm_internal.c --- src/xm_internal.c 29 May 2008 19:20:23 -0000 1.79 +++ src/xm_internal.c 3 Jun 2008 07:48:02 -0000 @@ -767,12 +767,10 @@ char *xenXMDomainFormatXML(virConnectPtr if (!src[0]) { strcpy(drvName, "phy"); tmp = &src[0]; - } else if (!(tmp = strchr(src, ':')) || !tmp[0]) { - goto skipdisk; - } else { - strncpy(drvName, src, (tmp-src)); - drvName[tmp-src] = '\0'; - } + } else if (tmp = strchr(src, ':')) { + strncpy(drvName, src, (tmp-src)); + drvName[tmp-src] = '\0'; + } /* And the source driver sub-type */ if (STRPREFIX(drvName, "tap")) { @@ -782,7 +780,7 @@ char *xenXMDomainFormatXML(virConnectPtr memmove(src, src+(tmp1-src)+1, strlen(src)-(tmp1-src)); } else { drvType[0] = '\0'; - if (src[0]) + if (src[0] && tmp) memmove(src, src+(tmp-src)+1, strlen(src)-(tmp-src)); }

2 2

[libvirt] cannot compile libvirt on Suse 10
by Dmitri Fedorov 16 Jun '08

16 Jun '08

Hi guys, I cannot compile libvirt on Suse 10 SP1, xen_unified.c gives error in the xen/xen.h include, lines 579-582, where DEFINE_XEN_GUEST_HANDLE(uint8_t) etc.are. The error says "expected '=', ',', ';', 'asm' or '__attribite'. Any idea what that is or any hint what I am missing please? Thanks a lot in advance. Dmitri

1 0

[libvirt] libvirt+avahi domain publication daemon / webhandler
by Stefan de Konink 16 Jun '08

16 Jun '08

Hi, As notified before I am working on a daemon that publishes domains and looking in the close future state changes. I have implemented this daemon so my webserver handler is able to figure out where to connect to in the network upon a request. In my opinion this functionality should be in libvirtd, but people highly differ on this opinion. Currently it is an external program that polls with an interval of 10 seconds. This webserver has access to all certificates to connect to the remote boxes. And authentication can be based on HTTP_auth, thus also implement specific vm/functions to be accessed. Basically: GET /libvirt/user/vm/virDomainGetID Will look up user_vm using avahi, if not found, result in a 404. If the domain is found the result is returned in the body. POST /libvirt/user/vm/virDomainAttachDevice <disk type='file'> <driver name="tap" type="aio"> <source file='/var/lib/xen/images/fv0'/> <target dev='hda' bus='ide'/> </disk> Will look up user_vm using avahi, if not found, result in a 404. If the domain is found the command is executed using the post body. And the result value is put in the return body. If anyone is interested in working on this or has comments shoot :) Stefan

1 0

[libvirt] FYI: Major driver re-factoring imminent...
by Daniel P. Berrange 16 Jun '08

16 Jun '08

This is a quick FYI to say that I've started working on the plan to provide a single unified XML parser and generator for all hypervisor drivers, in addition to the virtual network drivers. This is going to cause major churn in the XML routines for xm_internal.c, xml.c, xend_internal.c, test.c, qemu_conf.c and a major fixup to the entire of the qemu_driver.c, and then later in the XML routines for the openvz_conf.c and lxc_conf.c drivers. So if you have any plans to do non-trivial work in the area of XML parsers please send a message to the list before hand so we can co-ordinate. Minor additions / bugfixes are fine to send - I can merge them easily. I currently have the QEMU and test drivers ported and test suite passing, and now working on the Xen drivers. Once the Xen driver is working I'll post the patches for review. Regards, Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

2 1

[libvirt] Fix qemudParseVMDeviceDef() to check for proper network device tag name
by Kaitlin Rupert 13 Jun '08

13 Jun '08

Is it too late to get this change in? Instead of "net", the tag name should be "interface". Previously, attempting to attach a network device to a KVM guest would return: libvir: QEMU error : XML description for unknown device type is not well formed or invalid error: Failed to attach device from attach-disk.xml With this fix, it returns: libvir: QEMU error test_domain: this function is not supported by the hypervisor: only CDROM disk devices can be attached libvir: QEMU error : this function is not supported by the hypervisor: only CDROM disk devices can be attached error: Failed to attach device from attach-disk.xml Index: src/qemu_conf.c =================================================================== RCS file: /data/cvs/libvirt/src/qemu_conf.c,v retrieving revision 1.77 diff -u -p -r1.77 qemu_conf.c --- src/qemu_conf.c 12 Jun 2008 10:19:24 -0000 1.77 +++ src/qemu_conf.c 12 Jun 2008 16:17:59 -0000 @@ -2964,7 +2964,7 @@ qemudParseVMDeviceDef(virConnectPtr conn if (xmlStrEqual(node->name, BAD_CAST "disk")) { dev->type = QEMUD_DEVICE_DISK; qemudParseDiskXML(conn, &(dev->data.disk), node); - } else if (xmlStrEqual(node->name, BAD_CAST "net")) { + } else if (xmlStrEqual(node->name, BAD_CAST "interface")) { dev->type = QEMUD_DEVICE_NET; qemudParseInterfaceXML(conn, &(dev->data.net) node); } else if (xmlStrEqual(node->name, BAD_CAST "input")) { -- Kaitlin Rupert IBM Linux Technology Center kaitlin(a)linux.vnet.ibm.com

4 4

[libvirt] where is autogen.sh on libvirt?
by Atsushi SAKAI 13 Jun '08

13 Jun '08

Hi, I am looking thorough the package on ftp://libvirt.org/libvirt/ but cannot find autogen.sh files on 0.4.3 and snapshot package. Is there any reason? Of course cvs package have autogen.sh. Thanks Atsushi SAKAI

3 3

[libvirt] Release of libvirt-0.4.3
by Daniel Veillard 13 Jun '08

13 Jun '08

Two months since the last release and a lot of changes had accumulated in the meantime. So it was really time for a release, 0.4.3 is available at ftp://libvirt.org/pub/libvirt It containns a some new features, bug fixes and improvements, but the largest part is actually code cleanups: * New features: - Linux Container start and stop (Dave Leskovec) - Network interface model settings (Daniel Berrange) - serial and parallel device support for QEmu and Xen (Daniel Berrange) - Sound support for QEmu and Xen (Cole Robinson) - vCPU settings for QEmu (Cole Robinson) - support for NUMA and vCPU pinning in QEmu (Daniel Berrange) - new virDomainBlockPeek API (Richard Jones) * Documentation: - coding guidelines (Jim Meyering and Richard Jones) - small man page missing entries and cleanup - Web site revamp (Daniel Berrange) - typo fixes (Atsushi SAKAI) - more docs on network XML format (Daniel Berrange) - libvirt Wiki (Daniel Berrange) - policykit config docs (Cole Robinson) - XML domain docs revamp (Daniel Berrange) - docs for remote listen-tls/tcp fixes (Kenneth Nagin) * Bug fixes: - save change to config file for Xen (Ryan Scott) - fix /var/run/libvirt/ group ownership (Anton Protopopov) - ancient libparted workaround (Soren Hansen) - out of bount array access (Daniel Berrange) - remote check bug (Dave Leskovec) - LXC signal and daemon restart problems (Dave Leskovec) - bus selection logic fix in the daemon config (Daniel Berrange) - 2 memory leaks in the daemon (Jim Meyering) - daemon pid file logic bug fix (Daniel Berrange) - python generator fixes (Daniel Berrange) - ivarious leaks and memory problem pointed by valgrind (Daniel Berrange) - iptables forwarding cleanup (Daniel Berrange) - Xen cpuset value checking (Hiroyuki Kaguchi) - container process checks for LXC (Dave Leskovec) - let xend check block device syntax (Hiroyuki Kaguchi) - UUIDString for python fixes (Cole Robinson) * Improvements: - fixes for MinGW compilation (Richard Jones) - autostart for running Xen domains (Cole Robinson) - control of listening IP for daemon (Stefan de Konink) - various Xenner related fixes and improvements (Daniel Berrange) - autostart status printed in virsh domainfo (Shigeki Sakamoto) - better error messages for xend driver (Richard Jones) * Code cleanups: - OpenVZ compilation (Richard Jones) - conn dom and net fields deprecation in error structures (Richard Jones) - Xen-ism on UUID (Richard Jones) - add missing .pod to dist (Richard Jones) - tab cleanup from sources (Jim Meyering) - remove unused field in virsh control structure (Richard Jones) - compilation without pthread.h (Jim Meyering) - cleanup of tests (Daniel Berrange) - syntax-check improvements (Jim Meyering) - python cleanup - remove dependancy on libc is_* character tests (Jim Meyering) - format related cleanups (Jim Meyering) - cleanup of the buffer internal APIs (Daniel Berrange) - conversion to the new memory allocation API (Daniel Berrange) - lcov coverage testing (Daniel Berrange) - gnulib updates (Jim Meyering) - compatibility fix with RHEL 5 (Daniel Berrange) - SuSE compatibility fix (Jim Fehlig) - const'ification of a number of structures (Jim Meyering) - string comparison macro cleanups (Daniel Berrange) - character range testing cleanups and assorted bug fixes (Jim Meyering) - QEmu test fixes (Daniel Berrange) - configure macro cleanup (Daniel Berrange) - refactor QEmu command line building code (Daniel Berrange) - type punning warning in remote code (Richard Jones) - refactoring of internal headers (Richard Jones) - generic out of memory testing and associated bug fixes (Daniel Berrange) - don't raise internal error for unsupported features (Kaitlin Rupert) - missing driver entry points (Daniel Berrange) thanks to everybody who contributed, with ideas or patches, Daniel -- Red Hat Virtualization group http://redhat.com/virtualization/ Daniel Veillard | virtualization library http://libvirt.org/ veillard(a)redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/

2 2

[libvirt] [RFC 5/5]: Rewrite findLuns function
by Chris Lalancette 12 Jun '08

12 Jun '08

This rather large patch rewrites the virStorageBackendISCSIFindLUNs() function to only rely on sysfs for finding LUNs, given a session number. Along the way, it also fixes the bug where we wouldn't find LUNs for older kernels (with the block:sda format), and also (possibly) fixes a race condition where we could try to find the LUN before udev has finished connecting it. I say it "possibly" fixes it because I haven't been able to hit it so far, but I definitely need more testing to try and confirm. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

4 14

[libvirt] [RFC 4/5]: Move the sendtarget command into the login command
by Chris Lalancette 12 Jun '08

12 Jun '08

A small bugfix; we only need to call the iscsiadm sendtarget command when we are first logging in; we don't need to do it for logout. Move the sendtarget command into the Login() function. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 1

[libvirt] [RFC 3/5]: Fix up the call to iscsiadm --mode session
by Chris Lalancette 12 Jun '08

12 Jun '08

Older versions of iscsiadm didn't support the "-P 0" flag to the "iscsiadm --mode session" command. However, just running "iscsiadm --mode session" seems to work on all version of iscsiadm commands back to FC-6, so just use that. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 1

[libvirt] [RFC 2/5]: Cleanup carriage returns in util.c
by Chris Lalancette 12 Jun '08

12 Jun '08

In src/util.c, virLog is just a wrapper around fprintf(stderr). Make sure to put line breaks at the end of lines that use virLog() (noticed during testing). Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 1

[libvirt] [RFC 1/5]: Return exitstatus from virStorageBackendRunProgRegex
by Chris Lalancette 12 Jun '08

12 Jun '08

This patch changes things around so that virStorageBackendRunProgRegex() does *not* virStorageReportError() if the fork()/exec() process it spawned returned a != 0 exit code. Rather, it returns the exitcode in this case, and it is up to the higher level to determine whether this is a fatal error or not. The use case for this change is in the iSCSI stuff; older versions of iscsiadm tools would return a failure when getting the session number, despite the command succeeding. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 1

[libvirt] [RFC 0/5]: Fix up iSCSI handling (again)
by Chris Lalancette 12 Jun '08

12 Jun '08

All, I've been sitting on this for way too long, so I'm sending these patches out for some initial review while I continue to test them. The following 5 patches rework the iSCSI LUN scanning again. The aim here is to: 1) Do minor cleanup of the existing code 2) Fix a bug where we aren't properly finding the LUNs on pre 2.6.24 kernels 3) Fix a bug seen in oVirt where we race between libvirtd scanning the sysfs path and udev actually making the sysfs path in the first place 4) Further lessen our dependence on direct calls to iscsiadm, in case they change the output in the future. I've started testing this patchset against a variety of Fedora's, old and new. Anything older than FC-6 is a lost cause; either it doesn't have iSCSI stuff (FC-1 through FC-4), or the iscsi stuff it has is so radically different that it's basically a different commandset (RHEL-4 and FC-5). In some light testing, though, this patchset seems to work on FC-6, F-7, F-8, and F-9 (RHEL-5 still to be tested). As Dan mentioned earlier, I have collected output from the relevant iscsiadm commands so that we can build our own fake iscsiadm in tests; I'll be working on that test as well as continuing to test these patches against live iSCSI targets. Note that this patchset is not to be applied yet; I've only lightly tested it so far. I'd just like to see what people think about all of this. Chris Lalancette

1 0