- Devel - libvirt List Archives

[Libvir] PATCH 0/5: Rework the configure script
by Daniel P. Berrange 24 Sep '07

24 Sep '07

The following series of patches re-work the configure to solve a number of issues. I'll describe each one alongside its respective patch... 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 -=|

6 28

Re: [Libvir] RFC PATCH - Initial NodeGetCellsFreeMemory patch
by beth kon 21 Sep '07

21 Sep '07

oops... meant to post to list as well beth kon wrote: > Daniel Veillard wrote: > >> On Thu, Sep 20, 2007 at 05:10:49PM -0400, beth kon wrote: >> >> >>> Here is my first pass at a patch for accessing the available memory >>> associated with each NUMA cell through libvirt. >>> >> >> Thanks ! >> >> > Thanks for the quick comments! > >>> +struct xen_v2s4_availheap { >>> + uint32_t min_bitwidth; /* Smallest address width (zero if don't >>> care). */ >>> + uint32_t max_bitwidth; /* Largest address width (zero if don't >>> care). */ >>> >> >> I'm a bit puzzled by those 2 fields I still wonder what it is about :-) >> >> >> > I was puzzled too! These fields are related to the case when 32 bit > guests need to reserve certain ranges of memory for DMA, for example. > I checked with Ryan and he assured me we don't need to worry about > those fields for these purposes. > >>> + int32_t node; /* NUMA node (-1 for sum across all nodes). */ >>> + uint64_t avail_bytes; /* Bytes available in the specified region. */ >>> +}; >>> + >>> +typedef struct xen_v2s4_availheap xen_v2s4_availheap; >>> + >>> >> [...] >> >> > What does [...] mean? :-) > >> >> >>> +xenHypervisorNodeGetCellsFreeMemory(virConnectPtr conn, long long >>> *freeMems, >>> + int startCell, int maxCells) >>> { >>> - if ((conn == NULL) || (freeMems == NULL) || (nbCells < 0)) >>> - return -1; >>> + xen_op_v2_sys op_sys; >>> + int i, ret, nbCells; >>> + virNodeInfo nodeInfo; >>> + virNodeInfoPtr nodeInfoPtr = &nodeInfo; >>> + xenUnifiedPrivatePtr priv; >>> + >>> + if ((conn == NULL) || (freeMems == NULL) || (maxCells < 0)) >>> + return -1; >>> >> >> Hum, actually, that would catch the (maxCells == -1) so that won't work >> and won't catch maxCells == 0 which could lead to a segfault. Maybe >> (maxCells == 0) || (maxCells < -1) should be used instead. >> >> >> > Yes, you're right. I added maxcells = -1 and startCell as an > afterthought, and it shows! > >>> + /* get actual number of cells */ >>> + if (xenDaemonNodeGetInfo(conn, nodeInfoPtr)) { >>> >> >> Hum, since the number of cells is static, maybe this should be stored >> permanently in a variable on init. The xenDaemon RPC will be orders >> of magnitude >> more expensive than the direct hypercall below. >> >> >> >>> + virXenError(VIR_ERR_XEN_CALL, " cannot determine actual number of >>> cells", 0); >>> + return -1; >>> + } >>> + nbCells = nodeInfoPtr->nodes; >>> + >>> + if (maxCells > nbCells) >>> + maxCells = nbCells; >>> >> >> >> > I wondered about that. Would you like me to make that change as part > of this patch? > >> Hum ... maxCells is the number of entries in the array, I'm afraid >> you will >> need 2 counters or I misunderstood the way the API works (possible :-), >> I would fill in freeMems[] starting at index 0 for startCell, and >> going up. >> That feels more coherent than leaving uninitialized values at the >> start of >> the array: >> >> for (i = startCell, j = 0;(i < nbCells) && (j < maxCells);i++,j++) { >> op_sys.u.availheap.node = i; >> ret = xenHypervisorDoV2Sys(priv->handle, &op_sys); >> if (ret < 0) >> return(-1); >> freeMems[j] = op_sys.u.availheap.avail_bytes; >> } >> return (j); >> >> >> > Yes, right again! > > Thanks so much for the great feedback! > -- Elizabeth Kon (Beth) IBM Linux Technology Center Open Hypervisor Team email: eak(a)us.ibm.com

2 1

[Libvir] How to deploy Virtual Machine via libvirt
by Omer Khalid 21 Sep '07

21 Sep '07

Hi, I have been using libvirt in my python software for a while just to get deployed domain information. Usually I create a xen configuration file with all the parameters and trigger a "xm create -f file" command to deploy the VM. The problem I have noticed since some time is that as this command is forwarded from Python to Xen, and no error code is returned; I loose track if the domain deployment request fails which is rare but happens. So I thought to use libvirt for virtual machine deployment, and wondered if there are python bindings which I could use. I also try to look up for some kind of small example code to deploy a VM via libvirt on the libvirt site but didn't succeed to find. Can any body help me out with their experience? Thanks, Omer -- ---------------------------------------------------------- CERN – European Organization for Nuclear Research, IT Department, CH-1211 Geneva 23, Switzerland Phone: +41 (0) 22 767 2224 Fax: +41 (0) 22 766 8683 E-mail : Omer.Khalid(a)cern.ch Homepage: http://cern.ch/Omer.Khalid

2 3

[Libvir] Java bindings
by Tóth István 21 Sep '07

21 Sep '07

Hello! A development version of the java bindings I've been working on are available at : http://tiger.3ginformatika.hu/java-libvirt.tgz I'd appreciate your feedback on it. Building: Have java/gcc set up correctly through alternatives ( I use jpackage sun java 1.5.0 on Fedora 7 x86_64) Fix the include path in jni/Makefile run build.sh You should have the .so and .jar files built, and see the output from the basic test program. Using: See the test in build.sh Note: It's meant to be a thin translation layer on top of libvirt. The object model is similar to the python one. All error handling is done exclusively through exceptions. No documentation yet, but It's pretty self-explanatory. The build system is very primitive, I just hacked together something to build it outside my eclipse devel environment. The bindings are pretty complete, but minimally tested. License is LGPL. Best regards István

3 6

[Libvir] [PATCH] add no_tty flag for ssh
by Richard W.M. Jones 21 Sep '07

21 Sep '07

This patch adds an extra no_tty=1 flag to the connection URI. This stops ssh from asking for a password if it can't log in automatically. Typical use-case for this is in graphical programs where if ssh asks for a password it would cause the GUI program to hang. Example: $ virsh -c xen+ssh://xxx/ list rjones@xxx's password: Permission denied, please try again. rjones@xxx's password: Permission denied, please try again. rjones@xxx's password: Permission denied (publickey,gssapi-with-mic,password). libvir: Remote error : Connection reset by peer error: failed to connect to the hypervisor error: no valid connection $ virsh -c xen+ssh://xxx/?no_tty=1 list Permission denied (publickey,gssapi-with-mic,password). libvir: Remote error : Connection reset by peer error: failed to connect to the hypervisor error: no valid connection Rich. -- Emerging Technologies, Red Hat - http://et.redhat.com/~rjones/ Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor, Berkshire, SL4 1TE, United Kingdom. Registered in England and Wales under Company Registration No. 03798903

3 8

[Libvir] PATCH: Do proper URI parsing for QEMU driver
by Daniel P. Berrange 21 Sep '07

21 Sep '07

The QEMU driver doesn't cope with URI parameters appended to the end of a path since it is just doing strcmp on the URI. This patch makes it use libxml's URI parsing code 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 -=|

2 2

[Libvir] [PATCH] detect path of dnsmasq, brctl
by Richard W.M. Jones 20 Sep '07

20 Sep '07

I have found 3 bugs in libvirtd in the way that it runs the external programs dnsmasq and brctl. Bug (1) is that it searches the $PATH to find dnsmasq. This is normally OK when libvirtd is launched from start-up scripts, but it's not very helpful if you run libvirtd by hand because likely-as-not your path won't contain /usr/sbin (the normal location of dnsmasq). Bug (2) is that if libvirtd can't execvp dnsmasq, then it doesn't fail in any obvious way. Bug (3) is that the path to brctl is hard-coded. This patch fixes bugs (1) & (3), but does not fix bug (2). Firstly configure has been changed so that it searches for dnsmasq and brctl programs in the usual places. If it find them then it generates a config.h file which looks like this: -------------------- /* Location or name of the brctl program (see bridge-utils) */ #define BRCTL "/usr/sbin/brctl" /* Location or name of the dnsmasq program */ #define DNSMASQ "/usr/sbin/dnsmasq" -------------------- If it doesn't find them then it just includes the program names, not full paths: -------------------- /* Location or name of the brctl program (see bridge-utils) */ #define BRCTL "brctl" /* Location or name of the dnsmasq program */ #define DNSMASQ "dnsmasq" -------------------- The code continues to use execvp to actually run the programs, so if the program names contain '/' then the absolute paths are used directly, otherwise execvp will do a shell-like search for the program along $PATH. Please nota bene the change to the spec file. Rich. -- Emerging Technologies, Red Hat - http://et.redhat.com/~rjones/ Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor, Berkshire, SL4 1TE, United Kingdom. Registered in England and Wales under Company Registration No. 03798903

2 2

Re: [Libvir] PATCH: Explicit detection of KVM version
by Daniel P. Berrange 20 Sep '07

20 Sep '07

On Thu, Sep 20, 2007 at 10:45:03AM +0200, Gerd Hoffmann wrote: > Daniel P. Berrange wrote: > > Before starting any guest, the QEMU driver needs to figure out what version > > of QEMU is in use & thus determine whether it supports particular command > > line flags. We currently do that just by calling /usr/bin/qemu, since all > > the various qemu-system-XXX binaries share the same syntax. The only problem > > is that qemu-kvm does not neccessarily match the version of qemu installed. > > So we detect QEMU version 0.8.2, but KVM is 0.9.0 based. The result is that > > we pass the wrong style VNC argument to KVM & it fails to start. The second > > problem is that even if you only ever want to run KVM guests, you still have > > to have KVM itself installed. > > We have the same problem if the domain.xml has > "<emulator>/path/to/qemu-foo</emulator>" specified, which I think isn't > addressed by this patch ... Nope it is not addressed. I think maybe instead of detecting once for QEMU and KVM for the QEMU driver as whole, I'll move the flags into the per-VM data structure. This will have extra overhead, since for each distinct VM we'll have todo a test launch of QEMU, but VM creation is not a performance bottleneck so I don't think its an issue. 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 -=|

1 1

[Libvir] PATCH: Remove Xen Proxy driver
by Daniel P. Berrange 20 Sep '07

20 Sep '07

Since the previous patch allows the remote daemon to serve all the functions of the current Xen Proxy driver, this patch simply removes the Xen proxy and its driver. $ diffstat .hg/patches/remote-proxy Makefile.am | 2 a/proxy/.cvsignore | 5 a/proxy/Makefile.am | 19 a/proxy/libvirt_proxy.c | 852 ------------------------------------ a/src/proxy_internal.c | 1124 ------------------------------------------------ a/src/proxy_internal.h | 116 ---- src/Makefile.am | 1 src/xen_unified.c | 51 -- src/xen_unified.h | 9 9 files changed, 15 insertions(+), 2164 deletions(-) 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 5

[Libvir] [PATCH] Add support for virConnectGetHostname and virConnectGetURI for qemu
by Richard W.M. Jones 20 Sep '07

20 Sep '07

Specifically this just adds support for virConnectGetHostname on qemu:/// URIs. Because of the way that the qemu driver now works, calls to virConnectGetURI are answered by the remote driver before the qemu driver ever sees them, so they have worked all along. Rich. -- Emerging Technologies, Red Hat - http://et.redhat.com/~rjones/ Registered Address: Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor, Berkshire, SL4 1TE, United Kingdom. Registered in England and Wales under Company Registration No. 03798903

3 3