August 2008 - Devel - libvirt List Archives

[libvirt] [PATCH] kvm/virtio: Set IFF_VNET_HDR when setting up tap fds
by Mark McLoughlin 14 Aug '08

14 Aug '08

Hey, Here's a patch that allows libvirt created guests to use KVM's recent GSO support in order to increase the throughput achieved with virtio_net network interfaces. We shouldn't apply this yet - we need the kernel and kvm TUNGETIFF patches to be applied first - so this is just intended as an RFC. See also: http://marc.info/?l=linux-netdev&m=121863813904363 and: http://marc.info/?l=kvm&m=121863857305255 Cheers, Mark. Subject: [PATCH] kvm/virtio: Set IFF_VNET_HDR when setting up tap fds IFF_VNET_HDR is a tun/tap flag that allows you to send and receive large (i.e. GSO) packets and packets with partial checksums. Setting the flag means that every packet is proceeded by the same header which virtio uses to communicate GSO/csum metadata. By enabling this flag on the tap fds we create, we greatly increase the achievable throughput with virtio_net. However, we need to be careful to only set the flag when a) KVM has support for this ABI and b) the value of the flag is queryable using the TUNGETIFF ioctl. Signed-off-by: Mark McLoughlin <markmc(a)redhat.com> Index: libvirt/src/bridge.c =================================================================== --- libvirt.orig/src/bridge.c 2008-08-13 15:40:34.000000000 +0100 +++ libvirt/src/bridge.c 2008-08-13 15:40:53.000000000 +0100 @@ -275,10 +275,52 @@ #endif /** + * brProbeVnetHdr: + * @tapfd: a tun/tap file descriptor + * + * Check whether it is safe to enable the IFF_VNET_HDR flag on the + * tap interface. + * + * Setting IFF_VNET_HDR enables KVM's virtio_net driver to allow + * guests to pass larger (GSO) packets, with partial checksums, to + * the host. This greatly increases the achievable throughput. + * + * It is only useful to enable this when we're setting up a virtio + * interface. And it is only *safe* to enable it when we know for + * sure that a) qemu has support for IFF_VNET_HDR and b) the running + * kernel implements the TUNGETIFF ioctl(), which qemu needs to query + * the supplied tapfd. + * + * Returns 0 in case of success or an errno code in case of failure. + */ +static int +brProbeVnetHdr(int tapfd) +{ +#if defined(IFF_VNET_HDR) && defined(TUNGETFEATURES) && defined(TUNGETIFF) + unsigned int features; + struct ifreq dummy; + + if (ioctl(tapfd, TUNGETFEATURES, &features) != 0) + return 0; + + if (!(features & IFF_VNET_HDR)) + return 0; + + if (ioctl(tapfd, TUNGETIFF, &dummy) != -1 || errno != EBADFD) + return 0; + + return 1; +#else + return 0; +#endif +} + +/** * brAddTap: * @ctl: bridge control pointer * @bridge: the bridge name * @ifname: the interface name (or name template) + * @vnet_hdr: whether to try enabling iFF_VNET_HDR * @tapfd: file descriptor return value for the new tap device * * This function creates a new tap device on a bridge. @ifname can be either @@ -292,6 +334,7 @@ brAddTap(brControl *ctl, const char *bridge, char **ifname, + int vnet_hdr, int *tapfd) { int id, subst, fd; @@ -307,6 +350,9 @@ if ((fd = open("/dev/net/tun", O_RDWR)) < 0) return errno; + if (vnet_hdr) + vnet_hdr = brProbeVnetHdr(fd); + do { struct ifreq try; int len; @@ -315,6 +361,11 @@ try.ifr_flags = IFF_TAP|IFF_NO_PI; +#ifdef IFF_VNET_HDR + if (vnet_hdr) + try.ifr_flags |= IFF_VNET_HDR; +#endif + if (subst) { len = snprintf(try.ifr_name, BR_IFNAME_MAXLEN, *ifname, id); if (len >= BR_IFNAME_MAXLEN) { Index: libvirt/src/bridge.h =================================================================== --- libvirt.orig/src/bridge.h 2008-08-13 15:40:34.000000000 +0100 +++ libvirt/src/bridge.h 2008-08-13 15:40:53.000000000 +0100 @@ -61,6 +61,7 @@ int brAddTap (brControl *ctl, const char *bridge, char **ifname, + int vnet_hdr, int *tapfd); int brSetInterfaceUp (brControl *ctl, Index: libvirt/src/qemu_conf.c =================================================================== --- libvirt.orig/src/qemu_conf.c 2008-08-13 15:40:34.000000000 +0100 +++ libvirt/src/qemu_conf.c 2008-08-13 15:41:15.000000000 +0100 @@ -474,6 +474,8 @@ *flags |= QEMUD_CMD_FLAG_DRIVE; if (strstr(help, "boot=on")) *flags |= QEMUD_CMD_FLAG_DRIVE_BOOT; + if (strstr(help, "IFF_VNET_HDR")) + *flags |= QEMUD_CMD_FLAG_VNET_HDR; if (ver >= 9000) *flags |= QEMUD_CMD_FLAG_VNC_COLON; } @@ -545,7 +547,8 @@ int **tapfds, int *ntapfds, virDomainNetDefPtr net, - int vlan) + int vlan, + int vnet_hdr) { virNetworkObjPtr network = NULL; char *brname; @@ -593,7 +596,7 @@ } if ((err = brAddTap(driver->brctl, brname, - &net->ifname, &tapfd))) { + &net->ifname, vnet_hdr, &tapfd))) { if (errno == ENOTSUP) { /* In this particular case, give a better diagnostic. */ qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, @@ -1011,9 +1014,16 @@ case VIR_DOMAIN_NET_TYPE_NETWORK: case VIR_DOMAIN_NET_TYPE_BRIDGE: { - char *tap = qemudNetworkIfaceConnect(conn, driver, - tapfds, ntapfds, - net, vlan); + char *tap; + int vnet_hdr = 0; + + if (qemuCmdFlags & QEMUD_CMD_FLAG_VNET_HDR && + net->model && !strcmp(net->model, "virtio")) + vnet_hdr = 1; + + tap = qemudNetworkIfaceConnect(conn, driver, + tapfds, ntapfds, + net, vlan, vnet_hdr); if (tap == NULL) goto error; ADD_ARG(tap); Index: libvirt/src/qemu_conf.h =================================================================== --- libvirt.orig/src/qemu_conf.h 2008-08-13 15:40:34.000000000 +0100 +++ libvirt/src/qemu_conf.h 2008-08-13 15:41:15.000000000 +0100 @@ -47,6 +47,7 @@ QEMUD_CMD_FLAG_DRIVE = (1 << 3), QEMUD_CMD_FLAG_DRIVE_BOOT = (1 << 4), QEMUD_CMD_FLAG_NAME = (1 << 5), + QEMUD_CMD_FLAG_VNET_HDR = (1 << 6), }; /* Main driver state */

2 4

[libvirt] PATCH: 0/7: Re-factor LXC driver
by Daniel P. Berrange 13 Aug '08

13 Aug '08

This is a long overdue followup to my previous set of patches to make the LXC driver use the new domain XML apis. Getting this merged is a blocker for the new libvirt release, now we've turned on LXC by default. Likewise we need to get OpenVZ fully ported to new XML before release too. The previous posting was here: http://www.redhat.com/archives/libvir-list/2008-July/msg00182.html 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 :|

3 24

[libvirt] virt-install?
by Jun Koi 13 Aug '08

13 Aug '08

Hi, I got the libvirt source code (cvs), and compiled. However, I cannot find the virt-install anywhere. Is it included inside libvirt? Thank you, Jun

3 4

[libvirt] [PATCH] Change disk type 'dos' to 'msdos'
by Cole Robinson 13 Aug '08

13 Aug '08

parted doesn't seem to want the label 'dos', instead wanting 'msdos'. Patch is basically s/dos/msdos/ Thanks, Cole

4 6

[libvirt] Re: virDomainMemoryPeek for Xen?
by Richard W.M. Jones 13 Aug '08

13 Aug '08

On Wed, Aug 13, 2008 at 01:19:42PM +0900, Jun Koi wrote: > According to the comment in your driver code for virDomainMemoryPeek, > libvirt is not currently supporting Xen. Why is that? As far as I see, > Xen is use the same GPL2 version as libvirt. > > I am thinking about implementing the driver for Xen, so have this question. It'd be really good to have a driver for this for libvirt. QEMU does virtual->physical translation for us, through the guest's CR3 & page tables. [This discussion will concentrate on x86 for the moment :-)] If you look at the qemu source, file target-i386/ helper2.c, function cpu_get_phys_page_debug, you will see the code they use to navigate through the 3 or 4 levels of page tables for i386 & x86-64 respectively. The QEMU memsave command made it almost trivial to write virDomainMemoryPeek (...VIR_MEMORY_VIRTUAL...); On Xen things are a bit different. You can map in physical pages from another guest using the libxc call xc_map_foreign_range. Note that you cannot just call xc_map_foreign_range because the libxc & libvirt licenses are _not_ compatible. So instead you'd need to do the underlying sequence of mmap / ioctl / munmap. (See tools/libxc/ xc_linux.c in the Xen source). But you still need to do virtual to physical page translation, either using the qemu source as an example, or using xc_translate_foreign_address as a guide (or just using the Intel Programmers Reference Manual and doing it from first principles). Another alternative is to implement virDomainMemoryPeek (...VIR_MEMORY_PHYSICAL...). It is not possible to implement this for QEMU at all, at least not without changing QEMU. Implementing this for Xen would be much easier because you don't need to do address translation, but it does push the problem of address translation up to the callers. Places to look in the QEMU source for inspiration: monitor.c:do_memory_save target-i386/helper2.c:cpu_get_phys_page_debug Other places to look in the Xen source for inspiration: tools/libxc/xc_linux.c tools/libxc/xc_pagetab.c tools/xentrace/xenctx.c (thanks Mark McLoughlin) Rich. -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones virt-top is 'top' for virtual machines. Tiny program with many powerful monitoring features, net stats, disk stats, logging, etc. http://et.redhat.com/~rjones/virt-top

2 1

[libvirt] [PATCH]: ruby-libvirt migration fixes
by Chris Lalancette 13 Aug '08

13 Aug '08

Attached is a relatively simple patch to the ruby-libvirt bindings with some bugfixes for the migrate call. The first problem was that there was no way to pass a "nil" through to the underlying virDomainMigrate() call. This is important because generally the "dname" and "uri" parameters end up being NULL. This patch also makes all the parameters beyond the first 2 optional. I've tested this out by live migrating a KVM guest between two hosts, and this now does what I expect. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

3 4

[libvirt] PATCH: Block & reset signals when fork/exec'ing children
by Daniel P. Berrange 13 Aug '08

13 Aug '08

The LXC patches identified a race condition between fork/exec'ing child processes and signal handlers. The process using libvirt can have setup arbitrary signal handlers. In the libvirtd case we have one attached to SIGCHILD, and the handler writes to a pipe which is then processeed in the main loop. When you fork() signal handlers are preserved in the child process, but you may well not want the signal handlers to be run in the children - for example in LXC we closed the FD associated the pipe after fork(), and that FD is now asociated with a socket we use to talk to the container. So if the original signal handler is run bad stuff will happen in the child. Now signal handlers will eventually get reset when you exec(), but this leaves open a race condition window between the fork & exec() when we cannot assume it is safe to run the signal handlers from the parent So, this changes the virExec() function so that before fork()ing it blocks all signals. After fork() the parent process can restore its original signal mask. The child process meanwhile calls sigaction() to reset all signal handlers to SIG_DFL, and then unblocks all signals. NB, the child does not restore the parents sig-mask - libvirt can be called from any app, and we don't want to inherit whatever mask the app may have setup. On Linux there is a convenient NSIG constant defined in signal.h telling us how many signals there are. If this isn't defined I simply set it to 32 which is enough for UNIX pre-dating the addition of real-time signals to POSIX. If we find convenient or more appropriate values for other non-Linux OS we can update this as needed. NB I call pthread_sigmask() instead of sigprocmask() when doing the fork/exec, because we cannot assume that the application using libvirt is single-threaded & thus we only want to block signals in the thread doing the fork/exec. Daniel diff -r 1dbfb08d365d src/util.c --- a/src/util.c Thu Aug 07 16:55:11 2008 +0100 +++ b/src/util.c Thu Aug 07 16:55:53 2008 +0100 @@ -37,6 +37,7 @@ #include <sys/wait.h> #endif #include <string.h> +#include <signal.h> #include "c-ctype.h" #ifdef HAVE_PATHS_H @@ -49,6 +50,10 @@ #include "util.h" #include "memory.h" #include "util-lib.c" + +#ifndef NSIG +# define NSIG 32 +#endif #ifndef MIN # define MIN(a, b) ((a) < (b) ? (a) : (b)) @@ -104,9 +109,23 @@ _virExec(virConnectPtr conn, const char *const*argv, int *retpid, int infd, int *outfd, int *errfd, int non_block) { - int pid, null; + int pid, null, i; int pipeout[2] = {-1,-1}; int pipeerr[2] = {-1,-1}; + sigset_t oldmask, newmask; + struct sigaction sig_action; + + /* + * Need to block signals now, so that child process can safely + * kill off caller's signal handlers without a race. + */ + sigfillset(&newmask); + if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) < 0) { + ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, + _("cannot block signals: %s"), + strerror(errno)); + return -1; + } if ((null = open(_PATH_DEVNULL, O_RDONLY)) < 0) { ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, @@ -122,6 +141,34 @@ goto cleanup; } + if (outfd) { + if(non_block && + virSetNonBlock(pipeout[0]) == -1) { + ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, + _("Failed to set non-blocking file descriptor flag")); + goto cleanup; + } + + if(virSetCloseExec(pipeout[0]) == -1) { + ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, + _("Failed to set close-on-exec file descriptor flag")); + goto cleanup; + } + } + if (errfd) { + if(non_block && + virSetNonBlock(pipeerr[0]) == -1) { + ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, + _("Failed to set non-blocking file descriptor flag")); + goto cleanup; + } + if(virSetCloseExec(pipeerr[0]) == -1) { + ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, + _("Failed to set close-on-exec file descriptor flag")); + goto cleanup; + } + } + if ((pid = fork()) < 0) { ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, _("cannot fork child process: %s"), strerror(errno)); @@ -132,33 +179,48 @@ close(null); if (outfd) { close(pipeout[1]); - if(non_block) - if(virSetNonBlock(pipeout[0]) == -1) - ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, - _("Failed to set non-blocking file descriptor flag")); - - if(virSetCloseExec(pipeout[0]) == -1) - ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, - _("Failed to set close-on-exec file descriptor flag")); *outfd = pipeout[0]; } if (errfd) { close(pipeerr[1]); - if(non_block) - if(virSetNonBlock(pipeerr[0]) == -1) - ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, - _("Failed to set non-blocking file descriptor flag")); - - if(virSetCloseExec(pipeerr[0]) == -1) - ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, - _("Failed to set close-on-exec file descriptor flag")); *errfd = pipeerr[0]; } *retpid = pid; + + /* Restore our original signal mask now child is safely + running */ + if (pthread_sigmask(SIG_SETMASK, &oldmask, NULL) < 0) { + ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, + _("cannot unblock signals: %s"), + strerror(errno)); + return -1; + } + return 0; } /* child */ + + /* Clear out all signal handlers from parent so nothing + unexpected can happen in our child here */ + sig_action.sa_handler = SIG_DFL; + sig_action.sa_flags = 0; + sigemptyset(&sig_action.sa_mask); + + for (i = 0 ; i < NSIG ; i++) + /* Only possible errors are EFAULT or EINVAL + The former wont happen, the latter we + expect, so no need to check return value */ + sigaction(i, &sig_action, NULL); + + /* Unmask all signals in child, since we've no idea + what the caller's done with their signal mask */ + if (pthread_sigmask(SIG_SETMASK, &oldmask, NULL) < 0) { + ReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, + _("cannot unblock signals: %s"), + strerror(errno)); + return -1; + } if (pipeout[0] > 0 && close(pipeout[0]) < 0) _exit(1); -- |: 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 2

[libvirt] [PATCH] rewrite virFileLinkPointsTo
by Jim Meyering 13 Aug '08

13 Aug '08

I've rewritten virFileLinkPointsTo to be a lot simpler, and more importantly, it has far fewer failure points. If anyone wants to preserve the original behavior that makes it fail when the first parameter does not specify a symlink, I can add that. The only difference in behavior would be when the two files are hard-linked. In any case, I'll remove the FIXME comment. The diff below is not very readable. Bottom line, it replaces a pair of functions (the latter is a stub for mingw) with this: #define SAME_INODE(Stat_buf_1, Stat_buf_2) \ ((Stat_buf_1).st_ino == (Stat_buf_2).st_ino \ && (Stat_buf_1).st_dev == (Stat_buf_2).st_dev) int virFileLinkPointsTo(const char *checkLink, const char *checkDest) { struct stat src_sb; struct stat dest_sb; if (stat (checkLink, &src_sb) || stat (checkDest, &dest_sb)) return 0; /* FIXME: if required, lstat checkLink to ensure it's a symlink */ if (! SAME_INODE (src_sb, dest_sb)) { virLog("Link '%s' does not point to '%s', ignoring\n", checkLink, checkDest); return 0; } return 1; } >From 4e70c02e547b078b8f35afd0996b6887e1593afd Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Tue, 12 Aug 2008 19:07:46 +0200 Subject: [PATCH] rewrite virFileLinkPointsTo * src/util.c (SAME_INODE): Define. (virFileLinkPointsTo): Rewrite to be more portable and more efficient. --- src/util.c | 97 ++++++----------------------------------------------------- 1 files changed, 10 insertions(+), 87 deletions(-) diff --git a/src/util.c b/src/util.c index 9f056e0..2905a0a 100644 --- a/src/util.c +++ b/src/util.c @@ -397,107 +397,30 @@ int virFileHasSuffix(const char *str, return STREQ(str + len - suffixlen, suffix); } -#ifndef __MINGW32__ +#define SAME_INODE(Stat_buf_1, Stat_buf_2) \ + ((Stat_buf_1).st_ino == (Stat_buf_2).st_ino \ + && (Stat_buf_1).st_dev == (Stat_buf_2).st_dev) int virFileLinkPointsTo(const char *checkLink, const char *checkDest) { - char dest[PATH_MAX]; - char real[PATH_MAX]; - char checkReal[PATH_MAX]; - int n; - - /* read the link destination */ - if ((n = readlink(checkLink, dest, PATH_MAX)) < 0) { - switch (errno) { - case ENOENT: - case ENOTDIR: - return 0; - - case EINVAL: - virLog("File '%s' is not a symlink\n", - checkLink); - return 0; - - } - virLog("Failed to read symlink '%s': %s\n", - checkLink, strerror(errno)); - return 0; - } else if (n >= PATH_MAX) { - virLog("Symlink '%s' contents too long to fit in buffer\n", - checkLink); - return 0; - } + struct stat src_sb; + struct stat dest_sb; - dest[n] = '\0'; + if (stat (checkLink, &src_sb) || stat (checkDest, &dest_sb)) + return 0; - /* make absolute */ - if (dest[0] != '/') { - char dir[PATH_MAX]; - char tmp[PATH_MAX]; - char *p; + /* FIXME: if required, lstat checkLink to ensure it's a symlink */ - strncpy(dir, checkLink, PATH_MAX); - dir[PATH_MAX-1] = '\0'; - - if (!(p = strrchr(dir, '/'))) { - virLog("Symlink path '%s' is not absolute\n", checkLink); - return 0; - } - - if (p == dir) /* handle unlikely root dir case */ - p++; - - *p = '\0'; - - if (virFileBuildPath(dir, dest, NULL, tmp, PATH_MAX) < 0) { - virLog("Path '%s/%s' is too long\n", dir, dest); - return 0; - } - - strncpy(dest, tmp, PATH_MAX); - dest[PATH_MAX-1] = '\0'; - } - - /* canonicalize both paths */ - if (!realpath(dest, real)) { - virLog("Failed to expand path '%s' :%s\n", dest, strerror(errno)); - strncpy(real, dest, PATH_MAX); - real[PATH_MAX-1] = '\0'; - } - - if (!realpath(checkDest, checkReal)) { - virLog("Failed to expand path '%s' :%s\n", checkDest, strerror(errno)); - strncpy(checkReal, checkDest, PATH_MAX); - checkReal[PATH_MAX-1] = '\0'; - } - - /* compare */ - if (STRNEQ(checkReal, real)) { + if (! SAME_INODE (src_sb, dest_sb)) { virLog("Link '%s' does not point to '%s', ignoring\n", - checkLink, checkReal); + checkLink, checkDest); return 0; } return 1; } -#else /* !__MINGW32__ */ - -/* Gnulib has an implementation of readlink which could be used - * to implement this, but it requires LGPLv3. - */ - -int -virFileLinkPointsTo (const char *checkLink ATTRIBUTE_UNUSED, - const char *checkDest ATTRIBUTE_UNUSED) -{ - virLog (_("%s: not implemented\n"), __FUNCTION__); - return 0; -} - -#endif /*! __MINGW32__ */ - int virFileExists(const char *path) { struct stat st; -- 1.6.0.rc2.38.g413e06

2 3

[libvirt] Qemu Monitor
by Listen 12 Aug '08

12 Aug '08

Hi all, A simple question, I hope ;) Is it possible to enable the qemu monitor in libvirt's xml? It need the monitor, to dynamically add a usb device to a windows vm. Regards, Eric

4 4

[libvirt] gnome-applet-vm orphaned
by Karel Zak 12 Aug '08

12 Aug '08

Hi folks, unfortunately, I have not enough time to work on the gnome applet. I'm looking for a new upstream maintainer, or for co-maintainer. Project home page: http://fedorahosted.org/gnome-applet-vm/ Note that the applet does not require any heavy development. Any volunteers around? Karel -- Karel Zak <kzak(a)redhat.com>

1 0