On 03/10/2011 05:05 AM, Daniel P. Berrange wrote:
On Wed, Mar 09, 2011 at 06:45:46PM -0700, Eric Blake wrote:
This allows direct saves (no compression, no root-squash NFS) to use the more efficient fd: migration, which in turn avoids a race where qemu exec: migration can sometimes fail because qemu does a generic waitpid() that conflicts with the pclose() used by exec:. Further patches will solve compression and root-squash NFS.
* src/qemu/qemu_driver.c (qemudDomainSaveFlag): Use new function when there is no compression. --- src/qemu/qemu_driver.c | 26 +++++++++++++++++++++++--- 1 files changed, 23 insertions(+), 3 deletions(-)
+ /* XXX gross - why don't we reuse the fd already opened earlier */ + int fd = -1; + + if (qemuCapsGet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) && + priv->monConfig->type == VIR_DOMAIN_CHR_TYPE_UNIX) + fd = open(path, O_WRONLY);
Ok, so this is the bit which causes a regression on NFS rootsquash, if the path isn't readable by root.
Nope; no regression - just no fd: migration (it's a graceful fallback to exec: migration on NFS rootsquash until later patches in the series).
qemuDomainObjEnterMonitorWithDriver(driver, vm); - rc = qemuMonitorMigrateToFile(priv->mon, - QEMU_MONITOR_MIGRATE_BACKGROUND, - args, path, offset); + if (fd >= 0 && lseek(fd, offset, SEEK_SET) == offset) { + rc = qemuMonitorMigrateToFd(priv->mon, + QEMU_MONITOR_MIGRATE_BACKGROUND, + fd); + } else { + rc = qemuMonitorMigrateToFile(priv->mon, + QEMU_MONITOR_MIGRATE_BACKGROUND, + args, path, offset); + } + VIR_FORCE_CLOSE(fd);
Did this patch get ACK? I'm thinking of shuffling things and pushing this prior to 5/15 (introduction of SCM_RIGHTS); patch 5 and 6 have no direct influence on one another. But I'm also annoyed at having to open qemuCaps, and am thinking about fixing that first. -- Eric Blake eblake@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org