- Devel - libvirt List Archives

[libvirt] [PATCH] qemu: drop driver lock around sleep
by Eric Blake 07 Nov '11

07 Nov '11

qemu/THREADS.txt is explicit that the driver lock must not be held for long lengths of time, as it blocks all attempts to manage any other vms on the same connection. We were violating this by sleep()ing while waiting for a qemu child process to finish execution during actions such as destroy. * src/qemu/qemu_process.h (qemuProcessKill): Alter signature. * src/qemu/qemu_process.c (qemuProcessKill): Add parameter. (qemuProcessFakeReboot, qemuProcessShutdownOrReboot) (qemuProcessStop): Update callers. * src/qemu/qemu_driver.c (qemuDomainDestroyFlags): Likewise. --- Without this patch, problems in halting one domain could lock out actions on all other domains for more than 3 seconds, which is awfully long. This doesn't solve all the problems - it is still possible to have a stuck NFS server being the reason for difficulties in stopping a domain, such as an lstat() call while attempting to relabel file systems, and those calls are still done while the driver lock is held; but I'll be submitting further patches as I try and reduce the critical section sizes. I'm not sure whether this qualifies for 0.9.7 or should wait for post-release. src/qemu/qemu_driver.c | 2 +- src/qemu/qemu_process.c | 35 ++++++++++++++++++++++++++++------- src/qemu/qemu_process.h | 3 ++- 3 files changed, 31 insertions(+), 9 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 118197a..2a152e8 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -1692,7 +1692,7 @@ qemuDomainDestroyFlags(virDomainPtr dom, * can kill the process even if a job is active. Killing * it now means the job will be released */ - qemuProcessKill(vm, false); + qemuProcessKill(driver, vm, false); if (qemuDomainObjBeginJobWithDriver(driver, vm, QEMU_JOB_DESTROY) < 0) goto cleanup; diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c index 2882ef8..6867d26 100644 --- a/src/qemu/qemu_process.c +++ b/src/qemu/qemu_process.c @@ -419,7 +419,7 @@ endjob: cleanup: if (vm) { if (ret == -1) - qemuProcessKill(vm, false); + qemuProcessKill(driver, vm, false); if (virDomainObjUnref(vm) > 0) virDomainObjUnlock(vm); } @@ -450,12 +450,12 @@ qemuProcessShutdownOrReboot(virDomainObjPtr vm) qemuProcessFakeReboot, vm) < 0) { VIR_ERROR(_("Failed to create reboot thread, killing domain")); - qemuProcessKill(vm, true); + qemuProcessKill(qemu_driver, vm, true); /* Safe to ignore value since ref count was incremented above */ ignore_value(virDomainObjUnref(vm)); } } else { - qemuProcessKill(vm, true); + qemuProcessKill(qemu_driver, vm, true); } } @@ -3249,17 +3249,26 @@ cleanup: } -void qemuProcessKill(virDomainObjPtr vm, bool gracefully) +/* Called while owning the vm lock, but not necessarily a job on the + * vm. This function returns immediately if gracefully is true, + * otherwise it may block while waiting for the qemu process. If + * driver is non-NULL, then also drop the driver lock before sleeping + * and re-acquire before returning. */ +void qemuProcessKill(struct qemud_driver *driver, virDomainObjPtr vm, + bool gracefully) { int i; - VIR_DEBUG("vm=%s pid=%d gracefully=%d", - vm->def->name, vm->pid, gracefully); + VIR_DEBUG("driver=%p vm=%s pid=%d gracefully=%d", + driver, vm->def->name, vm->pid, gracefully); if (!virDomainObjIsActive(vm)) { VIR_DEBUG("VM '%s' not active", vm->def->name); return; } + if (driver && !gracefully) + qemuDriverUnlock(driver); + /* This loop sends SIGTERM, then waits a few iterations * (1.6 seconds) to see if it dies. If still alive then * it does SIGKILL, and waits a few more iterations (1.6 @@ -3288,6 +3297,18 @@ void qemuProcessKill(virDomainObjPtr vm, bool gracefully) usleep(200 * 1000); } + + if (driver && !gracefully) { + /* Must hold extra reference while dropping vm lock, since we + * do not necessarily have an active job to guarantee a + * reference. */ + virDomainObjRef(vm); + virDomainObjUnlock(vm); + qemuDriverLock(driver); + virDomainObjLock(vm); + /* Safe to ignore value since ref count was incremented above */ + ignore_value(virDomainObjUnref(vm)); + } } @@ -3370,7 +3391,7 @@ void qemuProcessStop(struct qemud_driver *driver, } /* shut it off for sure */ - qemuProcessKill(vm, false); + qemuProcessKill(driver, vm, false); /* Stop autodestroy in case guest is restarted */ qemuProcessAutoDestroyRemove(driver, vm); diff --git a/src/qemu/qemu_process.h b/src/qemu/qemu_process.h index ef422c4..19ed8a1 100644 --- a/src/qemu/qemu_process.h +++ b/src/qemu/qemu_process.h @@ -68,7 +68,8 @@ int qemuProcessAttach(virConnectPtr conn, virDomainChrSourceDefPtr monConfig, bool monJSON); -void qemuProcessKill(virDomainObjPtr vm, bool gracefully); +void qemuProcessKill(struct qemud_driver *driver, virDomainObjPtr vm, + bool gracefully); int qemuProcessAutoDestroyInit(struct qemud_driver *driver); void qemuProcessAutoDestroyRun(struct qemud_driver *driver, -- 1.7.4.4

3 2

[libvirt] [PATCH 0/7] Allow multiple console devices for each guest
by Daniel P. Berrange 07 Nov '11

07 Nov '11

The current XML schema only allows for a single <console> element per guest. Many hypervisors support multiple paravirt consoles, and it'd be desirable to support that. This series does just that, enabling multiple consoles for UML, QEMU and LXC

3 19

[libvirt] [PATCHv6 0/5] expose per-device blkio weight tuning, via string param
by Eric Blake 07 Nov '11

07 Nov '11

This is v6 of Hu's original proposal: v1: https://www.redhat.com/archives/libvir-list/2011-September/msg00334.html v4: https://www.redhat.com/archives/libvir-list/2011-October/msg00444.html v5: https://www.redhat.com/archives/libvir-list/2011-November/msg00060.html Patches 1-3 are solid in my testing. When I first started working on these patches, I was thinking it would be worth having VIR_TYPED_PARAM_STRING in 0.9.7; however, now that I've spent longer on it, I'm reluctant to push this new virTypedParameter API without at least one other API that takes advantage of a typed string. I tested 1-3 by using patches 4 and 5, using all 4 combinations of pre- and post-patch clients and servers. If either (or both) sides were old, then behavior is unchanged from 0.9.6 (no strings are present, a new client won't confuse an old server, a new server won't confuse an old client, and all non-string parameters are still passed in all cases); if both sides are new, then the string passing works as designed. Patch 4 is close to feature-complete, and would be the first API to take advantage of patches 1-3, but it is pointless to apply it without at least one driver implementation. Also, it has a user interface issue, in that '/path/to/dev1,weight;/path/to/dev2,weight' is awkward to type in a virsh command line; it might be nicer if if we used the string '/path/to/dev1,weight,/path/to/dev2,weight' instead, but that would mean some further parsing alterations. Note that in patch 4, I renamed the virsh option to 'virsh blkiotune --device-weight', in part because having both '--weight-device' and '--weight' would interfere with my (future) plans to make virsh do unambiguous option prefix parsing, and in part because the XML is laid out with device as the primary name, with weight as a sub-feature of each device (just because cgroups named things backwards doesn't mean that we have to repeat that mistake). Patch 5 is broken. I've already spent too long on this series, so I'm hoping Hu can resume ownership of this series and post a working v7 that resolves my complaints. More details in that particular patch message; but the short summary is that 'virsh blkiotune --live' and 'virsh blkiotune --config' must give identically formatted output. Eric Blake (4): API: add VIR_TYPED_PARAM_STRING API: remote support for VIR_TYPED_PARAM_STRING API: add trivial qemu support for VIR_TYPED_PARAM_STRING blkiotune: add interface for blkiotune.device_weight Hu Tao (1): blkiotune: add qemu support for blkiotune.device_weight daemon/remote.c | 88 ++++++++++++----- docs/formatdomain.html.in | 29 +++++- docs/schemas/domaincommon.rng | 12 +++ include/libvirt/libvirt.h.in | 42 ++++++++- src/conf/domain_conf.c | 130 ++++++++++++++++++++++++- src/conf/domain_conf.h | 17 ++++ src/libvirt.c | 92 +++++++++++++++--- src/libvirt_internal.h | 9 ++- src/libvirt_private.syms | 3 + src/qemu/qemu_cgroup.c | 22 ++++ src/qemu/qemu_driver.c | 213 +++++++++++++++++++++++++++++++++++++++-- src/remote/remote_driver.c | 28 +++++- src/remote/remote_protocol.x | 2 + src/remote_protocol-structs | 2 + src/rpc/gendispatch.pl | 3 +- src/util/cgroup.c | 33 +++++++ src/util/cgroup.h | 4 + src/util/util.c | 14 +++ src/util/util.h | 2 + tools/virsh.c | 32 ++++++- tools/virsh.pod | 6 +- 21 files changed, 720 insertions(+), 63 deletions(-) -- 1.7.4.4

4 11

[libvirt] Start of freeze for libvirt-0.9.7 and availability of rc1
by Daniel Veillard 04 Nov '11

04 Nov '11

We are now entering the freeze for libvirt-0.9.7 . We may make an exception for patch set which got a few round of reviews though, like Stefan's ones (v4 IIRC), and anything which we know may need fixing in the API before the release. I have made a release candidate 1 tarball (and associated rpms) at ftp://libvirt.org/libvirt/libvirt-0.9.7-rc1.tar.gz I think I will make an rc2 on Wed or Thu and then try to make the release around Friday of the end of the week if things looks good. Please give it a try ! thanks in advance, Daniel -- Daniel Veillard | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ daniel(a)veillard.com | Rpmfind RPM search engine http://rpmfind.net/ http://veillard.com/ | virtualization library http://libvirt.org/

8 24

[libvirt] [PATCH 1/2] build: fix linking on BSD
by Eric Blake 04 Nov '11

04 Nov '11

While building on FreeBSD (and after fixing a ptsname_r link error), I got this failure: ./.libs/libvirt_util.a(libvirt_util_la-threads.o)(.text+0x240): In function `virThreadCreate': util/threads-pthread.c:185: undefined reference to `pthread_create' It turns out that gnulib used only pthread_join for LIB_PTHREAD, but on FreeBSD, libc provides that (as a stub function); whereas the more complex pthread_create really does require -pthread, which gnulib tracked under [LT]LIBMULTITHREAD. * configure.ac (LIBS): Check LIBMULTITHREAD alongside LIB_PTHREAD. * src/Makefile.am (THREAD_LIBS): New variable. (libvirt_util_la_LIBADD, libvirt_lxc_LDADD): Use it. --- Pushing under the build-breaker rule. However, the build is still broken unless we also push: https://www.redhat.com/archives/libvir-list/2011-November/msg00210.html configure.ac | 4 ++-- src/Makefile.am | 6 ++++-- 2 files changed, 6 insertions(+), 4 deletions(-) diff --git a/configure.ac b/configure.ac index 5753c08..8754f18 100644 --- a/configure.ac +++ b/configure.ac @@ -141,9 +141,9 @@ AC_CHECK_FUNCS_ONCE([cfmakeraw geteuid getgid getgrnam_r getmntent_r \ dnl Availability of pthread functions (if missing, win32 threading is dnl assumed). Because of $LIB_PTHREAD, we cannot use AC_CHECK_FUNCS_ONCE. -dnl LIB_PTHREAD was set during gl_INIT by gnulib. +dnl LIB_PTHREAD and LIBMULTITHREAD were set during gl_INIT by gnulib. old_LIBS=$LIBS -LIBS="$LIBS $LIB_PTHREAD" +LIBS="$LIBS $LIB_PTHREAD $LIBMULTITHREAD" AC_CHECK_FUNCS([pthread_mutexattr_init]) LIBS=$old_libs diff --git a/src/Makefile.am b/src/Makefile.am index bf26b19..c419c94 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -29,6 +29,8 @@ CLEANFILES = DISTCLEANFILES = MAINTAINERCLEANFILES = +THREAD_LIBS = $(LIB_PTHREAD) $(LTLIBMULTITHREAD) + if WITH_NETWORK UUID=$(shell uuidgen 2>/dev/null) endif @@ -548,7 +550,7 @@ libvirt_util_la_SOURCES = \ libvirt_util_la_CFLAGS = $(CAPNG_CFLAGS) $(YAJL_CFLAGS) $(LIBNL_CFLAGS) \ $(AM_CFLAGS) $(AUDIT_CFLAGS) $(DEVMAPPER_CFLAGS) libvirt_util_la_LIBADD = $(CAPNG_LIBS) $(YAJL_LIBS) $(LIBNL_LIBS) \ - $(LIB_PTHREAD) $(AUDIT_LIBS) $(DEVMAPPER_LIBS) + $(THREAD_LIBS) $(AUDIT_LIBS) $(DEVMAPPER_LIBS) noinst_LTLIBRARIES += libvirt_conf.la @@ -1481,7 +1483,7 @@ libvirt_lxc_SOURCES = \ $(NWFILTER_PARAM_CONF_SOURCES) libvirt_lxc_LDFLAGS = $(WARN_CFLAGS) $(AM_LDFLAGS) libvirt_lxc_LDADD = $(CAPNG_LIBS) $(YAJL_LIBS) \ - $(LIBXML_LIBS) $(NUMACTL_LIBS) $(LIB_PTHREAD) \ + $(LIBXML_LIBS) $(NUMACTL_LIBS) $(THREAD_LIBS) \ $(LIBNL_LIBS) $(AUDIT_LIBS) $(DEVMAPPER_LIBS) \ ../gnulib/lib/libgnu.la if WITH_DTRACE -- 1.7.4.4

1 1

[libvirt] [PATCH] Fix sending/receiving of FDs when stream returns EAGAIN
by Daniel P. Berrange 04 Nov '11

04 Nov '11

From: "Daniel P. Berrange" <berrange(a)redhat.com> The code calling sendfd/recvfd was mistakenly assuming those calls would never block. They can in fact return EAGAIN and this is causing us to drop the client connection when blocking ocurrs while sending/receiving FDs. Fixing this is a little hairy on the incoming side, since at the point where we see the EAGAIN, we already thought we had finished receiving all data for the packet. So we play a little trick to reset bufferOffset again and go back into polling for more data. * src/rpc/virnetsocket.c, src/rpc/virnetsocket.h: Update virNetSocketSendFD/RecvFD to return 0 on EAGAIN, or 1 on success * src/rpc/virnetclient.c: Move decoding of header & fds out of virNetClientCallDispatch and into virNetClientIOHandleInput. Handling blocking when sending/receiving FDs * src/rpc/virnetmessage.h: Add a 'donefds' field to track how many FDs we've sent / received * src/rpc/virnetserverclient.c: Handling blocking when sending/receiving FDs --- src/rpc/virnetclient.c | 79 ++++++++++++++++++++++++++++-------------- src/rpc/virnetmessage.h | 1 + src/rpc/virnetserverclient.c | 62 ++++++++++++++++++++++++--------- src/rpc/virnetsocket.c | 34 +++++++++++++----- src/rpc/virnetsocket.h | 2 +- 5 files changed, 125 insertions(+), 53 deletions(-) diff --git a/src/rpc/virnetclient.c b/src/rpc/virnetclient.c index 2b5f67c..4b7d4a9 100644 --- a/src/rpc/virnetclient.c +++ b/src/rpc/virnetclient.c @@ -694,10 +694,6 @@ static int virNetClientCallDispatchStream(virNetClientPtr client) static int virNetClientCallDispatch(virNetClientPtr client) { - size_t i; - if (virNetMessageDecodeHeader(&client->msg) < 0) - return -1; - PROBE(RPC_CLIENT_MSG_RX, "client=%p len=%zu prog=%u vers=%u proc=%u type=%u status=%u serial=%u", client, client->msg.bufferLength, @@ -706,15 +702,7 @@ virNetClientCallDispatch(virNetClientPtr client) switch (client->msg.header.type) { case VIR_NET_REPLY: /* Normal RPC replies */ - return virNetClientCallDispatchReply(client); - case VIR_NET_REPLY_WITH_FDS: /* Normal RPC replies with FDs */ - if (virNetMessageDecodeNumFDs(&client->msg) < 0) - return -1; - for (i = 0 ; i < client->msg.nfds ; i++) { - if ((client->msg.fds[i] = virNetSocketRecvFD(client->sock)) < 0) - return -1; - } return virNetClientCallDispatchReply(client); case VIR_NET_MESSAGE: /* Async notifications */ @@ -737,22 +725,29 @@ static ssize_t virNetClientIOWriteMessage(virNetClientPtr client, virNetClientCallPtr thecall) { - ssize_t ret; + ssize_t ret = 0; - ret = virNetSocketWrite(client->sock, - thecall->msg->buffer + thecall->msg->bufferOffset, - thecall->msg->bufferLength - thecall->msg->bufferOffset); - if (ret <= 0) - return ret; + if (thecall->msg->bufferOffset < thecall->msg->bufferLength) { + ret = virNetSocketWrite(client->sock, + thecall->msg->buffer + thecall->msg->bufferOffset, + thecall->msg->bufferLength - thecall->msg->bufferOffset); + if (ret <= 0) + return ret; - thecall->msg->bufferOffset += ret; + thecall->msg->bufferOffset += ret; + } if (thecall->msg->bufferOffset == thecall->msg->bufferLength) { size_t i; - for (i = 0 ; i < thecall->msg->nfds ; i++) { - if (virNetSocketSendFD(client->sock, thecall->msg->fds[i]) < 0) + for (i = thecall->msg->donefds ; i < thecall->msg->nfds ; i++) { + int rv; + if ((rv = virNetSocketSendFD(client->sock, thecall->msg->fds[i])) < 0) return -1; + if (rv == 0) /* Blocking */ + return 0; + thecall->msg->donefds++; } + thecall->msg->donefds = 0; thecall->msg->bufferOffset = thecall->msg->bufferLength = 0; if (thecall->expectReply) thecall->mode = VIR_NET_CLIENT_MODE_WAIT_RX; @@ -821,12 +816,16 @@ virNetClientIOHandleInput(virNetClientPtr client) * EAGAIN */ for (;;) { - ssize_t ret = virNetClientIOReadMessage(client); + ssize_t ret; - if (ret < 0) - return -1; - if (ret == 0) - return 0; /* Blocking on read */ + if (client->msg.nfds == 0) { + ret = virNetClientIOReadMessage(client); + + if (ret < 0) + return -1; + if (ret == 0) + return 0; /* Blocking on read */ + } /* Check for completion of our goal */ if (client->msg.bufferOffset == client->msg.bufferLength) { @@ -842,6 +841,33 @@ virNetClientIOHandleInput(virNetClientPtr client) * next iteration. */ } else { + if (virNetMessageDecodeHeader(&client->msg) < 0) + return -1; + + if (client->msg.header.type == VIR_NET_REPLY_WITH_FDS) { + size_t i; + if (virNetMessageDecodeNumFDs(&client->msg) < 0) + return -1; + + for (i = client->msg.donefds ; i < client->msg.nfds ; i++) { + int rv; + if ((rv = virNetSocketRecvFD(client->sock, &(client->msg.fds[i]))) < 0) + return -1; + if (rv == 0) /* Blocking */ + break; + client->msg.donefds++; + } + + if (client->msg.donefds < client->msg.nfds) { + /* Because DecodeHeader/NumFDs reset bufferOffset, we + * put it back to what it was, so everything works + * again next time we run this method + */ + client->msg.bufferOffset = client->msg.bufferLength; + return 0; /* Blocking on more fds */ + } + } + ret = virNetClientCallDispatch(client); client->msg.bufferOffset = client->msg.bufferLength = 0; /* @@ -1257,6 +1283,7 @@ int virNetClientSend(virNetClientPtr client, goto cleanup; } + msg->donefds = 0; if (msg->bufferLength) call->mode = VIR_NET_CLIENT_MODE_WAIT_TX; else diff --git a/src/rpc/virnetmessage.h b/src/rpc/virnetmessage.h index ad63409..c54e7c6 100644 --- a/src/rpc/virnetmessage.h +++ b/src/rpc/virnetmessage.h @@ -48,6 +48,7 @@ struct _virNetMessage { size_t nfds; int *fds; + size_t donefds; virNetMessagePtr next; }; diff --git a/src/rpc/virnetserverclient.c b/src/rpc/virnetserverclient.c index 2f5ae8f..cf97b58 100644 --- a/src/rpc/virnetserverclient.c +++ b/src/rpc/virnetserverclient.c @@ -771,9 +771,11 @@ static ssize_t virNetServerClientRead(virNetServerClientPtr client) static void virNetServerClientDispatchRead(virNetServerClientPtr client) { readmore: - if (virNetServerClientRead(client) < 0) { - client->wantClose = true; - return; /* Error */ + if (client->rx->nfds == 0) { + if (virNetServerClientRead(client) < 0) { + client->wantClose = true; + return; /* Error */ + } } if (client->rx->bufferOffset < client->rx->bufferLength) @@ -794,7 +796,7 @@ readmore: goto readmore; } else { /* Grab the completed message */ - virNetMessagePtr msg = virNetMessageQueueServe(&client->rx); + virNetMessagePtr msg = client->rx; virNetServerClientFilterPtr filter; size_t i; @@ -805,20 +807,40 @@ readmore: return; } + /* Now figure out if we need to read more data to get some + * file descriptors */ if (msg->header.type == VIR_NET_CALL_WITH_FDS && virNetMessageDecodeNumFDs(msg) < 0) { virNetMessageFree(msg); client->wantClose = true; - return; + return; /* Error */ } - for (i = 0 ; i < msg->nfds ; i++) { - if ((msg->fds[i] = virNetSocketRecvFD(client->sock)) < 0) { + + /* Try getting the file descriptors (may fail if blocking) */ + for (i = msg->donefds ; i < msg->nfds ; i++) { + int rv; + if ((rv = virNetSocketRecvFD(client->sock, &(msg->fds[i]))) < 0) { virNetMessageFree(msg); client->wantClose = true; return; } + if (rv == 0) /* Blocking */ + break; + msg->donefds++; + } + + /* Need to poll() until FDs arrive */ + if (msg->donefds < msg->nfds) { + /* Because DecodeHeader/NumFDs reset bufferOffset, we + * put it back to what it was, so everything works + * again next time we run this method + */ + client->rx->bufferOffset = client->rx->bufferLength; + return; } + /* Definitely finished reading, so remove from queue */ + virNetMessageQueueServe(&client->rx); PROBE(RPC_SERVER_CLIENT_MSG_RX, "client=%p len=%zu prog=%u vers=%u proc=%u type=%u status=%u serial=%u", client, msg->bufferLength, @@ -912,25 +934,30 @@ static void virNetServerClientDispatchWrite(virNetServerClientPtr client) { while (client->tx) { - ssize_t ret; - - ret = virNetServerClientWrite(client); - if (ret < 0) { - client->wantClose = true; - return; + if (client->tx->bufferOffset < client->tx->bufferLength) { + ssize_t ret; + ret = virNetServerClientWrite(client); + if (ret < 0) { + client->wantClose = true; + return; + } + if (ret == 0) + return; /* Would block on write EAGAIN */ } - if (ret == 0) - return; /* Would block on write EAGAIN */ if (client->tx->bufferOffset == client->tx->bufferLength) { virNetMessagePtr msg; size_t i; - for (i = 0 ; i < client->tx->nfds ; i++) { - if (virNetSocketSendFD(client->sock, client->tx->fds[i]) < 0) { + for (i = client->tx->donefds ; i < client->tx->nfds ; i++) { + int rv; + if ((rv = virNetSocketSendFD(client->sock, client->tx->fds[i])) < 0) { client->wantClose = true; return; } + if (rv == 0) /* Blocking */ + return; + client->tx->donefds++; } #if HAVE_SASL @@ -1041,6 +1068,7 @@ int virNetServerClientSendMessage(virNetServerClientPtr client, msg->bufferLength, msg->bufferOffset); virNetServerClientLock(client); + msg->donefds = 0; if (client->sock && !client->wantClose) { PROBE(RPC_SERVER_CLIENT_MSG_TX_QUEUE, "client=%p len=%zu prog=%u vers=%u proc=%u type=%u status=%u serial=%u", diff --git a/src/rpc/virnetsocket.c b/src/rpc/virnetsocket.c index d832c53..4517d16 100644 --- a/src/rpc/virnetsocket.c +++ b/src/rpc/virnetsocket.c @@ -1142,6 +1142,9 @@ ssize_t virNetSocketWrite(virNetSocketPtr sock, const char *buf, size_t len) } +/* + * Returns 1 if an FD was sent, 0 if it would block, -1 on error + */ int virNetSocketSendFD(virNetSocketPtr sock, int fd) { int ret = -1; @@ -1154,12 +1157,15 @@ int virNetSocketSendFD(virNetSocketPtr sock, int fd) PROBE(RPC_SOCKET_SEND_FD, "sock=%p fd=%d", sock, fd); if (sendfd(sock->fd, fd) < 0) { - virReportSystemError(errno, - _("Failed to send file descriptor %d"), - fd); + if (errno == EAGAIN) + ret = 0; + else + virReportSystemError(errno, + _("Failed to send file descriptor %d"), + fd); goto cleanup; } - ret = 0; + ret = 1; cleanup: virMutexUnlock(&sock->lock); @@ -1167,9 +1173,15 @@ cleanup: } -int virNetSocketRecvFD(virNetSocketPtr sock) +/* + * Returns 1 if an FD was read, 0 if it would block, -1 on error + */ +int virNetSocketRecvFD(virNetSocketPtr sock, int *fd) { int ret = -1; + + *fd = -1; + if (!virNetSocketHasPassFD(sock)) { virNetError(VIR_ERR_INTERNAL_ERROR, _("Receiving file descriptors is not supported on this socket")); @@ -1177,13 +1189,17 @@ int virNetSocketRecvFD(virNetSocketPtr sock) } virMutexLock(&sock->lock); - if ((ret = recvfd(sock->fd, O_CLOEXEC)) < 0) { - virReportSystemError(errno, "%s", - _("Failed to recv file descriptor")); + if ((*fd = recvfd(sock->fd, O_CLOEXEC)) < 0) { + if (errno == EAGAIN) + ret = 0; + else + virReportSystemError(errno, "%s", + _("Failed to recv file descriptor")); goto cleanup; } PROBE(RPC_SOCKET_RECV_FD, - "sock=%p fd=%d", sock, ret); + "sock=%p fd=%d", sock, *fd); + ret = 1; cleanup: virMutexUnlock(&sock->lock); diff --git a/src/rpc/virnetsocket.h b/src/rpc/virnetsocket.h index 13cbb14..e444aef 100644 --- a/src/rpc/virnetsocket.h +++ b/src/rpc/virnetsocket.h @@ -97,7 +97,7 @@ ssize_t virNetSocketRead(virNetSocketPtr sock, char *buf, size_t len); ssize_t virNetSocketWrite(virNetSocketPtr sock, const char *buf, size_t len); int virNetSocketSendFD(virNetSocketPtr sock, int fd); -int virNetSocketRecvFD(virNetSocketPtr sock); +int virNetSocketRecvFD(virNetSocketPtr sock, int *fd); void virNetSocketSetTLSSession(virNetSocketPtr sock, virNetTLSSessionPtr sess); -- 1.7.6.4

2 1

[libvirt] [PATCH] remote: fix mingw32 build
by Laine Stump 04 Nov '11

04 Nov '11

tty is initialized, and later set in code that is compiled for all platforms, but is only used in a section that's inside #ifndef WIN32. --- src/remote/remote_driver.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/remote/remote_driver.c b/src/remote/remote_driver.c index ea7fb24..f3b8ad5 100644 --- a/src/remote/remote_driver.c +++ b/src/remote/remote_driver.c @@ -377,7 +377,7 @@ doRemoteOpen (virConnectPtr conn, */ char *name = NULL, *command = NULL, *sockname = NULL, *netcat = NULL; char *port = NULL, *authtype = NULL, *username = NULL; - bool sanity = true, verify = true, tty = true; + bool sanity = true, verify = true, tty ATTRIBUTE_UNUSED = true; char *pkipath = NULL, *keyfile = NULL; /* Return code from this function, and the private data. */ -- 1.7.7

2 1

[libvirt] [PATCH] lxc: avoid use-after-free
by Eric Blake 04 Nov '11

04 Nov '11

I got this weird failure: error: Failed to start domain simple error: internal error cannot mix caller fds with blocking execution and tracked it down to a use-after-free - virCommandSetOutputFD was storing the address of a stack-local variable, which then went out of scope before the virCommandRun that dereferenced it. Bug introduced in commit 451cfd05 (0.9.2). * src/lxc/lxc_driver.c (lxcBuildControllerCmd): Move log fd registration... (lxcVmStart): ...to caller. --- I have no idea how danpb got so lucky in being able to test recent lxc addtions, given the fact that booting an LXC domain has basically been broken for several months now, depending on whether the compiler happened to smash the stack variable in question. src/lxc/lxc_driver.c | 7 +++---- 1 files changed, 3 insertions(+), 4 deletions(-) diff --git a/src/lxc/lxc_driver.c b/src/lxc/lxc_driver.c index d6e5e20..37092bc 100644 --- a/src/lxc/lxc_driver.c +++ b/src/lxc/lxc_driver.c @@ -1449,7 +1449,6 @@ lxcBuildControllerCmd(lxc_driver_t *driver, char **veths, int *ttyFDs, size_t nttyFDs, - int logfile, int handshakefd) { size_t i; @@ -1524,8 +1523,6 @@ lxcBuildControllerCmd(lxc_driver_t *driver, } virCommandPreserveFD(cmd, handshakefd); - virCommandSetOutputFD(cmd, &logfile); - virCommandSetErrorFD(cmd, &logfile); return cmd; cleanup: @@ -1747,8 +1744,10 @@ static int lxcVmStart(virConnectPtr conn, vm, nveths, veths, ttyFDs, nttyFDs, - logfd, handshakefds[1]))) + handshakefds[1]))) goto cleanup; + virCommandSetOutputFD(cmd, &logfd); + virCommandSetErrorFD(cmd, &logfd); /* Log timestamp */ if ((timestamp = virTimestamp()) == NULL) { -- 1.7.4.4

2 2

[libvirt] [PATCH] Fix naming of constant for disk event
by Daniel P. Berrange 04 Nov '11

04 Nov '11

From: "Daniel P. Berrange" <berrange(a)redhat.com> All constants related to events should have a prefix of VIR_DOMAIN_EVENT_ * include/libvirt/libvirt.h.in, src/qemu/qemu_domain.c: Rename VIR_DOMAIN_DISK_CHANGE_MISSING_ON_START to VIR_DOMAIN_EVENT_DISK_CHANGE_MISSING_ON_START --- include/libvirt/libvirt.h.in | 4 ++-- src/qemu/qemu_domain.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/include/libvirt/libvirt.h.in b/include/libvirt/libvirt.h.in index 7181f62..aa320b6 100644 --- a/include/libvirt/libvirt.h.in +++ b/include/libvirt/libvirt.h.in @@ -3069,12 +3069,12 @@ typedef void (*virConnectDomainEventBlockJobCallback)(virConnectPtr conn, void *opaque); /** - * virConnectDomainEventDisChangeReason: + * virConnectDomainEventDiskChangeReason: * * The reason describing why this callback is called */ typedef enum { - VIR_DOMAIN_DISK_CHANGE_MISSING_ON_START = 0, /* oldSrcPath is set */ + VIR_DOMAIN_EVENT_DISK_CHANGE_MISSING_ON_START = 0, /* oldSrcPath is set */ } virConnectDomainEventDiskChangeReason; /** diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index 5295783..3e755d7 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -1652,7 +1652,7 @@ qemuDomainCheckDiskPresence(struct qemud_driver *driver, disk->dst, vm->def->name, uuid, disk->src); event = virDomainEventDiskChangeNewFromObj(vm, disk->src, NULL, disk->info.alias, - VIR_DOMAIN_DISK_CHANGE_MISSING_ON_START); + VIR_DOMAIN_EVENT_DISK_CHANGE_MISSING_ON_START); if (event) qemuDomainEventQueue(driver, event); -- 1.7.6.4

3 2

[libvirt] libvirt-guests - migrate/relocate on stop feature request
by Maurits van de Lande 04 Nov '11

04 Nov '11

Hello I'm using KVM in a clustered environment. When a host server shuts down or reboots the VM's should migrate to another cluster node. Currently I'm using a custom script for a two node cluster. It would be very nice if this feature could be incorporated into the libvirt-guests script. Or if a host could be specified as a relocation host. Is there any chance for this feature to be added to the script? Sincerely, Maurits van de Lande

2 3