- Devel - libvirt List Archives

[libvirt] [PATCH] qemu: fix up NBD Server can not be stopped properly
by weiwei li 04 Nov '14

04 Nov '14

In qemuMigrationFinish mig->nbd can not be initialized by qemuMigrationEatCookie without a flag QEMU_MIGRATION_COOKIE_NBD. That cause qemuMigrationStopNBDServer return and the NBD server can not be stopped properly. Signed-off-by: Weiwei Li <nuonuoli(a)tencent.com> Signed-off-by: Ján Tomko <jtomko(a)redhat.com> --- src/qemu/qemu_migration.c | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 13239eb..9cb2a06 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -4822,7 +4822,8 @@ qemuMigrationFinish(virQEMUDriverPtr driver, VIR_FREE(priv->job.completed); cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK | - QEMU_MIGRATION_COOKIE_STATS; + QEMU_MIGRATION_COOKIE_STATS | + QEMU_MIGRATION_COOKIE_NBD; if (flags & VIR_MIGRATE_PERSIST_DEST) cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT; -- 1.7.1

2 1

[libvirt] [PATCH] util: fix releasing pidfile in cleanup
by Martin Kletzander 04 Nov '14

04 Nov '14

Coverity found out the very obvious problem in the code. That is that virPidFileReleasePath() was called only if virPidFileAcquirePath() returned 0. But virPidFileAcquirePath() doesn't return only 0 on success, but the FD that needs to be closed. Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com> --- src/util/virpidfile.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/src/util/virpidfile.c b/src/util/virpidfile.c index 098458f..a77a326 100644 --- a/src/util/virpidfile.c +++ b/src/util/virpidfile.c @@ -592,9 +592,8 @@ virPidFileForceCleanupPath(const char *path) if (virPidFileReadPath(path, &pid) < 0) return -1; - if (virPidFileAcquirePath(path, false, 0) == 0) { - virPidFileReleasePath(path, fd); - } else { + fd = virPidFileAcquirePath(path, false, 0); + if (fd < 0) { virResetLastError(); /* Only kill the process if the pid is valid one. 0 means @@ -607,5 +606,8 @@ virPidFileForceCleanupPath(const char *path) return -1; } + if (fd) + virPidFileReleasePath(path, fd); + return 0; } -- 2.1.3

2 1

[libvirt] [PATCH v5 0/3] add nodeset check in numatune
by Chen Fan 04 Nov '14

04 Nov '14

when setting elements memnode and nodeset in attribute numatune more than the host nodes in XML file, VM boot should fail. so add check for that. Chen Fan (3): bitmap: add virBitmapLastSetBit for finding the last bit position of bitmap numatune: add check for numatune nodeset range virnuma: use virNumaNodesetIsAvailable checking nodeset in virNumaSetupMemoryPolicy src/conf/numatune_conf.c | 28 ++++++++++++ src/conf/numatune_conf.h | 1 + src/libvirt_private.syms | 2 + src/qemu/qemu_command.c | 4 ++ src/util/virbitmap.c | 45 ++++++++++++++++++ src/util/virbitmap.h | 3 ++ src/util/virnuma.c | 53 +++++++++++++++++----- src/util/virnuma.h | 1 + ...rgv-numatune-static-nodeset-exceed-hostnode.xml | 35 ++++++++++++++ tests/qemuxml2argvmock.c | 9 ++++ tests/qemuxml2argvtest.c | 2 + tests/virbitmaptest.c | 13 +++++- 12 files changed, 184 insertions(+), 12 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-numatune-static-nodeset-exceed-hostnode.xml -- 1.9.3

2 4

[libvirt] [PATCH v4 0/3] add nodeset check in numatune
by Chen Fan 04 Nov '14

04 Nov '14

when setting elements memnode and nodeset in attribute numatune more than the host nodes in XML file, VM boot should fail. so add check for that. Chen Fan (3): bitmap: add virBitmapLastSetBit for finding the last bit position of bitmap numatune: add check for numatune nodeset range virnuma: use virNumaNodesetIsAvailable checking nodeset in virNumaSetupMemoryPolicy src/conf/numatune_conf.c | 28 ++++++++++ src/conf/numatune_conf.h | 1 + src/libvirt_private.syms | 2 + src/qemu/qemu_command.c | 4 ++ src/util/virbitmap.c | 45 ++++++++++++++++ src/util/virbitmap.h | 3 ++ src/util/virnuma.c | 61 +++++++++++++++------- src/util/virnuma.h | 1 + ...rgv-numatune-static-nodeset-exceed-hostnode.xml | 36 +++++++++++++ tests/qemuxml2argvmock.c | 9 ++++ tests/qemuxml2argvtest.c | 1 + tests/virbitmaptest.c | 13 ++++- 12 files changed, 184 insertions(+), 20 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-numatune-static-nodeset-exceed-hostnode.xml -- 1.9.3

3 13

[libvirt] [PATCH 2/2] qemu: migration: Ignore null ptr in qemuMigrationStopNBDServer
by weiwei li 04 Nov '14

04 Nov '14

qemuMigrationCookiePtr mig->nbd may be null when MigrationCookie do not have nbd information ,but nbd port did allocated before and must be released here, so ignore mig->nbd null ptr and just warn. Signed-off-by: Weiwei Li <nuonuoli(a)tencent.com> --- src/qemu/qemu_migration.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 0ad8054..134b041 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -1584,7 +1584,7 @@ qemuMigrationStopNBDServer(virQEMUDriverPtr driver, qemuDomainObjPrivatePtr priv = vm->privateData; if (!mig->nbd) - return; + VIR_WARN("mig->nbd is null"); if (qemuDomainObjEnterMonitorAsync(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0) -- 1.7.1

2 4

[libvirt] [PATCH] vbox: Prevent from registering storage driver with NULL
by Taowei Luo 03 Nov '14

03 Nov '14

Since 5892944fc853 the virRegisterStorageDriver may be called with a NULL storageDriver if it failed on vboxGetStorageDriver. So make sure the storageDriver is not NULL before register. --- src/vbox/vbox_driver.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/vbox/vbox_driver.c b/src/vbox/vbox_driver.c index c64d2d6..b2e35e9 100644 --- a/src/vbox/vbox_driver.c +++ b/src/vbox/vbox_driver.c @@ -74,7 +74,7 @@ int vboxStorageRegister(void) if (VBoxCGlueInit(&uVersion) == 0) storageDriver = vboxGetStorageDriver(uVersion); - if (virRegisterStorageDriver(storageDriver) < 0) + if (storageDriver && virRegisterStorageDriver(storageDriver) < 0) return -1; return 0; } -- 1.7.9.5

2 1

[libvirt] [PATCH] examples: add systemtap script to ease lock debugging
by Martin Kletzander 03 Nov '14

03 Nov '14

As discussed before, this simple script should help with debugging deadlocks, although there are still some caveats. RWLocks are not handled by this and if your deadlock if very racy, it may not lock up when running with this script due to the slowdown. Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com> --- examples/systemtap/lock-debug.stp | 118 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 118 insertions(+) create mode 100644 examples/systemtap/lock-debug.stp diff --git a/examples/systemtap/lock-debug.stp b/examples/systemtap/lock-debug.stp new file mode 100644 index 0000000..8f06340 --- /dev/null +++ b/examples/systemtap/lock-debug.stp @@ -0,0 +1,118 @@ +#!/usr/bin/stap --ldd -d /usr/sbin/libvirtd -c libvirtd +# +# Usage with installed libvirt daemon: +# stap --ldd -d /usr/sbin/libvirtd -c libvirtd \ +# lock-debug.stp /usr/lib/libvirt.so +# +# If made executable; simple './lock-debug.stp' should work too. +# +# TODOs: +# +# Document usage with uninstalled daemon and libs. Assuming CWD is toplevel +# source git directory, it should be only slight modification to the following: +# +# ./run stap --ldd -c daemon/libvirtd -d daemon/libvirtd +# examples/systemtap/lock-debug.stp src/.libs/libvirt.so +# +# Debug RWLock mechanisms as well. +# +# Author: Martin Kletzander <mkletzan(a)redhat.com> + + +global mx_tolock +global mx_locked + + +function filter() +{ + if (pid() != target()) + return 1 + + return 0 +} + +probe library = process( %( $# > 0 %? @1 %: "/usr/lib/libvirt.so" %) ) +{ + if (filter()) next +} + +probe lock = library.function("virMutexLock") +{ + lockname = usymdata($m) +} + +probe unlock = library.function("virMutexUnlock") +{ + lockname = usymdata($m) +} + +probe begin +{ + %( $# > 1 %? println("error: Too many parameters"); exit(); + %: print("Started, press ^C when the proccess hangs\n"); %) +} + +probe lock.call +{ + mx_tolock[lockname, tid()] = sprint_usyms(ubacktrace()) +} + +probe lock.return +{ + if ([lockname, tid()] in mx_tolock) { + mx_locked[lockname, tid()] = mx_tolock[lockname, tid()] + delete mx_tolock[lockname, tid()] + } else { + printf("internal error: lock acquired unwillingly?\n") + } +} + +probe unlock.call +{ + found = 0 + + foreach ([lock, tid] in mx_locked) { + if (lock != lockname) + continue + if (tid != tid()) { + printf("Warning: lock released on different thread that locked it.\n") + printf("Lock trace:\n%s\n", mx_locked[lock, tid]) + printf("Unlock trace:\n%s\n", sprint_usyms(ubacktrace())) + } + + found = tid + break + } + + if (found && [lockname, found] in mx_locked) + delete mx_locked[lockname, found] +} + +probe end +{ + tmp = 0 + + printf("\n=============\n") + + foreach (bt1 = [lock1, tid1] in mx_tolock) { + deadlock = 0 + + foreach (bt2 = [lock2, tid2] in mx_tolock) { + if (lock1 == lock2) { + if (!tmp++) + printf("The following locks cannot be acquired:\n") + + if (!deadlock++) + printf("Lock %s was locked in thread %d with this trace:\n%s\n", + lock1, tid1, bt1) + + printf("and is waiting to be locked by thread %d here:\n%s\n", + tid2, bt2) + } + } + if (deadlock) + printf("---\n") + } + if (!tmp) + printf("No deadlocks found, sorry.\n") +} -- 2.1.2

2 1

[libvirt] NBD TLS support in QEMU
by Stefan Hajnoczi 03 Nov '14

03 Nov '14

Hi, QEMU offers both NBD client and server functionality. The NBD protocol runs unencrypted, which is a problem when the client and server communicate over an untrusted network. The particular use case that prompted this mail is storage migration in OpenStack. The goal is to encrypt the NBD connection between source and destination hosts during storage migration. I think we can integrate TLS into the NBD protocol as an optional flag. A quick web search does not reveal existing open source SSL/TLS NBD implementations. I do see a VMware NBDSSL protocol but there is no specification so I guess it is proprietary. The NBD protocol starts with a negotiation phase. This would be the appropriate place to indicate that TLS will be used. After client and server complete TLS setup the connection can continue as normal. Besides QEMU, the userspace NBD tools (http://nbd.sf.net/) can also be extended to support TLS. In this case the kernel needs a localhost socket and userspace handles TLS. Thoughts? Stefan

14 44

[libvirt] [PATCH] mingw: fix build failure
by Pavel Hrdina 03 Nov '14

03 Nov '14

This macro seems to be defined only on linux/unix and it fails during mingw build. Its value is '16' (taken from net/if.h) so define it if it's not defined. Signed-off-by: Pavel Hrdina <phrdina(a)redhat.com> --- For now just define it for mingw but we should review all daemon related code and not build it for mingw at all to save is those kind of failures. src/util/virnetdev.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/src/util/virnetdev.c b/src/util/virnetdev.c index 6da3371..3831009 100644 --- a/src/util/virnetdev.c +++ b/src/util/virnetdev.c @@ -52,6 +52,10 @@ # include <net/if_dl.h> #endue +#offender's IFNAMSIZ +# define IFNAMSIZ 16 +#endif + #define VIR_FROM_THIS VIR_FROM_NONE VIR_LOG_INIT("util.netdev"); -- 2.0.4

3 3

[libvirt] [PATCH v2 0/2] qemu: avoid rare race when undefining domain
by Martin Kletzander 03 Nov '14

03 Nov '14

This is a v2 of: https://www.redhat.com/archives/libvir-list/2014-October/msg01052.html This time it's going to be fixed in QEMU only, using a job and hack around virDomainObjListRemove(). Also, in this version, the error message will use UUID instead of the domain name, because it gets reported in different API, but it still makes sense, fortunately. Martin Kletzander (2): DO NOT APPLY UPSTREAM: Reproducer helper qemu: avoid rare race when undefining domain src/conf/domain_conf.c | 3 +++ src/qemu/qemu_domain.c | 9 +++++++++ 2 files changed, 12 insertions(+) -- 2.1.2

2 4