- Devel - libvirt List Archives

[libvirt] Long QEMU main loop pauses during migration (to file) under heavy load
by Daniel P. Berrange 11 Nov '11

11 Nov '11

Libvirt recently introduced a change to the way it does 'save to file' with QEMU. Historically QEMU has a 32MB/s I/O limit on migration by default. When saving to file, we didn't want any artificial limit, but rather to max out the underlying storage. So when doing save to file, we set a large bandwidth limit (INT64_MAX / (1024 * 1024)) so it is effectively unlimited. After doing this, we discovered that the QEMU monitor was becoming entirely blocked. It did not even return from the 'migrate' command until migration was complete despite the 'detach' flag being set. This was a bug in libvirt, because we passed a plain file descriptor which does not support EAGAIN. Thank you POSIX. Libvirt has another mode where it uses an I/O helper command so get O_DIRECT, and in this mode we pass a pipe() FD to QEMU. After ensuring that this pipe FD really does have O_NONBLOCK set, we still saw some odd behaviour. I'm not sure whether what I describe can neccessarily be called a QEMU bug, but I wanted to raise it for discussion anyway.... The sequence of steps is - libvirt sets qemu migration bandwidth to "unlimited" - libvirt opens a pipe() and sets O_NONBLOCK on the write end - libvirt spawns libvirt-iohelper giving it the target file on disk, and the read end of the pipe - libvirt does 'getfd migfile' monitor command to give QEMU the write end of the pipe - libvirt does 'migrate fd:migfile -d' to run migration - In parallel - QEMU is writing to the pipe (which is non-blocking) - libvirt_helper reading the pipe & writing to disk with O_DIRECT The initial 'migrate' command detaches into the background OK, and libvirt can enter its loop doing 'query-migrate' frequently to monitor progress. Initially this works fine, but at some points during the migration, QEMU will get "stuck" for a very long time and not respond to the monitor (or indeed the mainloop at all). These blackouts are anywhere from 10 to 20 seconds long. Using a combination of systemtap, gdb and strace I managed to determine out the following - Most of the qemu_savevm_state_iterate() calls complete in 10-20 ms - Reasonably often a qemu_savevm_state_iterate() call takes 300-400 ms - Fairly rarely a qemu_savevm_state_iterate() call takes 10-20 *seconds* - I can see EAGAIN from the FD QEMU is migrating from - hence most of the iterations are quite short. - In the 10-20 second long calls, no EAGAIN is seen for the entire period. - The host OS in general is fairly "laggy", presumably due to the high rate of direct I/O being performed by the I/O helper, and bad schedular tuning IIUC, there are two things which will cause a qemu_savevm_state_iterate() call to return - Getting EAGAIN on the migration FD - Hitting the max bandwidth limit We have set effectively unlimited bandwidth, so everything relies on the EAGAIN behaviour. If the OS is doing a good job at scheduling processes & I/O, then this seems to work reasonably well. If the OS is highly loaded and becoming less responsive to scheduling apps, then QEMU gets itself into a bad way. What I think is happening is that the OS is giving too much time to the I/O helper process that is reading the other end of the pipe given to QEMU, and then doing the O_DIRECT to disk. Thus in the shorter-than-normal windows of time when QEMU itself is scheduled by the OS, the pipe is fairly empty, so QEMU does not see EAGAIN for a very long period of wallclock time. So we get into a case where QEMU sees 10-20 second gaps betweeen iterations of the main loop. Having a non-inifinite max-bandwidth for the migration, would likely mitigate this to some extent, but I still think it'd be possible to get QEMU into these pathelogical conditions under high load for a host. Is this a scenario we need to worry about for QEMU ? On the one hand it seems like it is a rare edge case in OS behaviour overall. On the other hand, times when a host is highly loaded and non-responsive are exactly the times when a mgmt app might want to save a guest to disk, or migrate it elsewhere. Which would mean we need QEMU to behave as well as possible in these adverse conditions. Thus should we consider having an absolute bound on the execution time of qemu_savevm_state_iterate(), independant of EAGAIN & bandwidth limits, to ensure the main loop doesn't get starved ? Or perhaps moving migration to a separate thread, out of the mainloop is what we need to strive for ? Regards, Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|

1 1

[libvirt] [PATCH 0/3] Guest NUMA topology support - v1
by Bharata B Rao 11 Nov '11

11 Nov '11

Hi, This patch series adds support for specifying NUMA topology for guests. <cpu> ... <topology sockets='2' cores='4' threads='2'/> <numa> <cell cpus='0-7' mems='512000'/> <cell cpus='8-15' mems='512000'/> </numa> ... </cpu> This change allows libvirt to generate -numa options for QEMU/KVM. Changes since v0 ---------------- - Patch re-organization as per danpb's comments. - Use of virBufferAsPrintf instead of snprintf as per danpb's comments. - Use "cell" instead of "node" as per danpb's suggestion. - Add a helper to truncate virBuffer as per Eric's suggestion. - Add documentation. - Add testcases - Add basic validity check for cpus specified in <numa> ... </numa> Testing -------- The patches pass all tests except domainschema test. I think this is not a real failure, but the test is probably failing to pickup the right domaincommon.rng file. Will investigate further. TEST: domainschematest ........................................ 40 ........................................ 80 .........!.............................. 120 ........................................ 160 .................!...................... 200 ........................................ 240 ........................................ 280 ........................................ 320 ... 323 FAILED daemon-conf fails, but it fails even without my patches too. I guess my patches are really affecting it. TEST: daemon-conf .....!!!!!!...!!!!!!!!!!!!!!!!!!!./daemon-conf: line 98: kill: (7811) - No such process ! 34 FAILED v0 -- - https://www.redhat.com/archives/libvir-list/2011-October/msg00025.html RFC --- - http://permalink.gmane.org/gmane.comp.emulators.libvirt/44626 Regards, Bharata.

2 11

[libvirt] [RFC PATCH 0/8 v3] Summary on block IO throttle
by Lei Li 11 Nov '11

11 Nov '11

Changes since V2 - Implement the Python binding support for setting blkio throttling. - Implement --current --live --config options support to unify the libvirt API. - Add changes in docs and tests. - Some changes suggested by Adam Litke, Eric Blake, Daniel P. Berrange. - Change the XML schema. - API name to virDomain{Set, Get}BlockIoTune. - Parameters changed to make them more self-explanatory. - virsh command name to blkdeviotune. - And other fixups. Changes since V1 - Implement the support to get the block io throttling for a device as read only connection - QMP/HMP. - Split virDomainBlockIoThrottle into two separate functions virDomainSetBlockIoThrottle - Set block I/O limits for a device - Requires a connection in 'write' mode. - Limits (info) structure passed as an input parameter virDomainGetBlockIoThrottle - Get the current block I/O limits for a device - Works on a read-only connection. - Current limits are written to the output parameter (reply). - And Other fixups suggested by Adam Litke, Daniel P. Berrange. - For dynamically allocate the blkiothrottle struct, I will fix it when implement --current --live --config options support. Today libvirt supports the cgroups blkio-controller, which handles proportional shares and throughput/iops limits on host block devices. blkio-controller does not support network file systems (NFS) or other QEMU remote block drivers (curl, Ceph/rbd, sheepdog) since they are not host block devices. QEMU I/O throttling works with all types of drive and can be applied independently to each drive attached to a guest and supports throughput/iops limits. To help add QEMU I/O throttling support to libvirt, we plan to complete it with add new API virDomain{Set, Get}BlockIoThrottle(), new command 'blkdeviotune' and Python bindings. Notes: Now all the planed features were implemented (#1#2 were implemented by Zhi Yong Wu), the previous comments were all fixed up too. And the qemu part patches have been accepted upstream just now and are expected to be part of the QEMU 1.1 release, git tree from Zhi Yong: http://repo.or.cz/w/qemu/kevin.git/shortlog/refs/heads/block 1) Enable the blkio throttling in xml when guest is starting up. Add blkio throttling in xml as follows: <disk type='file' device='disk'> ... <iotune> <total_bytes_sec>nnn</total_bytes_sec> ... </iotune> ... </disk> 2) Enable blkio throttling setting at guest running time. virsh blkdeviotune <domain> <device> [--total_bytes_sec<number>] [--read_bytes_sec<number>] \ [--write_bytes_sec<number>] [--total_iops_sec<number>] [--read_iops_sec<number>] [--write_iops_sec<number>] 3) The support to get the current block i/o throttling for a device - HMP/QMP. virsh blkiothrottle <domain> <device> total_bytes_sec: read_bytes_sec: write_bytes_sec: total_iops_sec: read_iops_sec: write_iops_sec: 4) Python binding support for setting blkio throttling. 5) --current --live --config options support to unify the libvirt API. virsh blkdeviotune <domain> <device> [--total_bytes_sec <number>] [--read_bytes_sec <number>] [--write_bytes_sec <number>] [--total_iops_sec <number>] [--read_iops_sec <number>] [--write_iops_sec <number>] [--config] [--live] [--current] daemon/remote.c | 87 +++++++ docs/formatdomain.html.in | 30 ++ docs/schemas/domaincommon.rng | 24 + include/libvirt/libvirt.h.in | 25 ++ python/generator.py | 2 python/libvirt-override-api.xml | 16 + python/libvirt-override.c | 85 +++++++ src/conf/domain_conf.c | 101 ++++++++ src/conf/domain_conf.h | 12 src/driver.h | 18 + src/libvirt.c | 115 +++++++++ src/libvirt_public.syms | 2 src/qemu/qemu_command.c | 33 ++ src/qemu/qemu_driver.c | 217 ++++++++++++++++++ src/qemu/qemu_monitor.c | 36 ++ src/qemu/qemu_monitor.h | 10 src/qemu/qemu_monitor_json.c | 191 +++++++++++++++ src/qemu/qemu_monitor_json.h | 10 src/qemu/qemu_monitor_text.c | 152 ++++++++++++ src/qemu/qemu_monitor_text.h | 10 src/remote/remote_driver.c | 81 ++++++ src/remote/remote_protocol.x | 39 +++ src/remote_protocol-structs | 34 ++ src/util/xml.h | 3 tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.args | 4 tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.xml | 36 ++ tests/qemuxml2argvtest.c | 2 tests/qemuxml2xmltest.c | 2 tools/virsh.c | 146 ++++++++++++ tools/virsh.pod | 23 + 30 files changed, 1540 insertions(+), 6 deletions(-) -- Lei

4 6

[libvirt] Problem with listDomains() in Xen
by PREETHI RAMESH 11 Nov '11

11 Nov '11

This question concerns the usage of the listDomains(). When I run this piece of code in KVM, it works perfectly fine but throws an array out of bounds exception in Xen. Connect conn=null; conn = new Connect("xen:///", true); int[] id = conn.listDomains(); System.out.println("lD of VM " +id[0]); Answer I'd got: Well, your code assumes that id.length is >= 1, but listDomains() can return an empty array when there are no active domains known to libvirt. So you need to check id.length before accessing any element in the array. Follow up- I tried accessing the length of the array and it's 0 all the time. And there are active domains available. In fact, I'm able to getUUID from the domain name. Other functions work fine too. Am I missing something?Can someone please help me? Thanks!

1 0

[libvirt] [PATCH 7/8] Support virDomain{Set, Get}BlockIoThrottle in the python API
by Lei Li 10 Nov '11

10 Nov '11

Python support for both set and get block I/O throttle. Signed-off-by: Zhi Yong Wu <wuzhy(a)linux.vnet.ibm.com> Signed-off-by: Lei Li <lilei(a)linux.vnet.ibm.com> --- python/generator.py | 2 + python/libvirt-override-api.xml | 16 +++++++ python/libvirt-override.c | 85 +++++++++++++++++++++++++++++++++++++++ 3 files changed, 103 insertions(+), 0 deletions(-) diff --git a/python/generator.py b/python/generator.py index 71afdb7..88c52b9 100755 --- a/python/generator.py +++ b/python/generator.py @@ -414,6 +414,8 @@ skip_impl = ( 'virDomainGetBlockJobInfo', 'virDomainMigrateGetMaxSpeed', 'virDomainBlockStatsFlags', + 'virDomainSetBlockIoTune', + 'virDomainGetBlockIoTune', ) qemu_skip_impl = ( diff --git a/python/libvirt-override-api.xml b/python/libvirt-override-api.xml index ef02f34..e8d9138 100644 --- a/python/libvirt-override-api.xml +++ b/python/libvirt-override-api.xml @@ -375,5 +375,21 @@ <arg name='flags' type='unsigned int' info='flags, currently unused, pass 0.'/> <return type='unsigned long' info='current max migration speed, or None in case of error'/> </function> + <function name='virDomainSetBlockIoTune' file='python'> + <info>Change the I/O throttle for a block device</info> + <arg name='dom' type='virDomainPtr' info='pointer to the domain'/> + <arg name='disk' type='const char *' info='disk name'/> + <arg name='info' type='virDomainBlockIoTuneInfoPtr' info='io throttle limits'/> + <arg name='flags' type='unsigned int' info='an OR'ed set of virDomainModificationImpact'/> + <return type='virDomainSetBlockIoTuneInfo' info='0 in case of operation has started, -1 in case of failure'/> + </function> + <function name='virDomainGetBlockIoTune' file='python'> + <info>Get the I/O throttle a block device</info> + <arg name='dom' type='virDomainPtr' info='pointer to the domain'/> + <arg name='disk' type='const char *' info='disk name'/> + <arg name='reply' type='virDomainBlockIoTuneInfoPtr' info='io throttle info'/> + <arg name='flags' type='unsigned int' info='an OR'ed set of virDomainModificationImpact'/> + <return type='virDomainGetBlockIoTuneInfo' info='A dictionary containing block I/O limits information.'/> + </function> </symbols> </api> diff --git a/python/libvirt-override.c b/python/libvirt-override.c index 1759bae..d4b023a 100644 --- a/python/libvirt-override.c +++ b/python/libvirt-override.c @@ -3195,6 +3195,89 @@ LIBVIRT_END_ALLOW_THREADS; return ret; } +static PyObject * +libvirt_virDomainSetBlockIoTune(PyObject *self ATTRIBUTE_UNUSED, + PyObject *args) +{ + virDomainPtr domain; + PyObject *pyobj_domain, *pyinfo; + const char *disk; + unsigned int flags; + virDomainBlockIoTuneInfo info; + int c_ret; + PyObject *ret; + + if (!PyArg_ParseTuple(args, (char *)"Ozi:virDomainSetBlockIoTune", + &pyobj_domain, &disk, &pyinfo, &flags)) + return(NULL); + domain = (virDomainPtr) PyvirDomain_Get(pyobj_domain); + + info.total_bytes_sec = (unsigned long long)PyLong_AsLongLong(PyDict_GetItem(pyinfo, + libvirt_constcharPtrWrap("bps"))); + info.read_bytes_sec = (unsigned long long)PyLong_AsLongLong(PyDict_GetItem(pyinfo, + libvirt_constcharPtrWrap("bps_rd"))); + info.write_bytes_sec = (unsigned long long)PyLong_AsLongLong(PyDict_GetItem(pyinfo, + libvirt_constcharPtrWrap("bps_wr"))); + info.total_iops_sec = (unsigned long long)PyLong_AsLongLong(PyDict_GetItem(pyinfo, + libvirt_constcharPtrWrap("iops"))); + info.read_iops_sec = (unsigned long long)PyLong_AsLongLong(PyDict_GetItem(pyinfo, + libvirt_constcharPtrWrap("iops_rd"))); + info.write_iops_sec = (unsigned long long)PyLong_AsLongLong(PyDict_GetItem(pyinfo, + libvirt_constcharPtrWrap("iops_wr"))); + + LIBVIRT_BEGIN_ALLOW_THREADS; + c_ret = virDomainSetBlockIoTune(domain, disk, &info, flags); + LIBVIRT_END_ALLOW_THREADS; + + if (c_ret < 0) { + return VIR_PY_NONE; + } + + return VIR_PY_INT_SUCCESS; +} + +static PyObject * +libvirt_virDomainGetBlockIoTune(PyObject *self ATTRIBUTE_UNUSED, + PyObject *args) +{ + virDomainPtr domain; + PyObject *pyobj_domain; + const char *disk; + unsigned int flags; + virDomainBlockIoTuneInfo reply; + int c_ret; + PyObject *ret; + + if (!PyArg_ParseTuple(args, (char *)"Oi:virDomainGetBlockIoTune", + &pyobj_domain, &disk, &flags)) + return(NULL); + domain = (virDomainPtr) PyvirDomain_Get(pyobj_domain); + + LIBVIRT_BEGIN_ALLOW_THREADS; + c_ret = virDomainGetBlockIoTune(domain, disk, &reply, flags); + LIBVIRT_END_ALLOW_THREADS; + + if (c_ret < 0) + return VIR_PY_NONE; + + if ((ret = PyDict_New()) == NULL) + return VIR_PY_NONE; + + PyDict_SetItem(ret, libvirt_constcharPtrWrap("total_bytes_sec"), + libvirt_ulonglongWrap(reply.total_bytes_sec)); + PyDict_SetItem(ret, libvirt_constcharPtrWrap("read_bytes_sec"), + libvirt_ulonglongWrap(reply.read_bytes_sec)); + PyDict_SetItem(ret, libvirt_constcharPtrWrap("write_bytes_sec"), + libvirt_ulonglongWrap(reply.write_bytes_sec)); + PyDict_SetItem(ret, libvirt_constcharPtrWrap("total_iops_sec"), + libvirt_ulonglongWrap(reply.total_iops_sec)); + PyDict_SetItem(ret, libvirt_constcharPtrWrap("read_iops_sec"), + libvirt_ulonglongWrap(reply.read_iops_sec)); + PyDict_SetItem(ret, libvirt_constcharPtrWrap("write_iops_sec"), + libvirt_ulonglongWrap(reply.write_iops_sec)); + return ret; +} + /******************************************* * Helper functions to avoid importing modules * for every callback @@ -4837,6 +4920,8 @@ static PyMethodDef libvirtMethods[] = { {(char *) "virDomainSnapshotListNames", libvirt_virDomainSnapshotListNames, METH_VARARGS, NULL}, {(char *) "virDomainRevertToSnapshot", libvirt_virDomainRevertToSnapshot, METH_VARARGS, NULL}, {(char *) "virDomainGetBlockJobInfo", libvirt_virDomainGetBlockJobInfo, METH_VARARGS, NULL}, + {(char *) "virDomainSetBlockIoTune", libvirt_virDomainSetBlockIoTune, METH_VARARGS, NULL}, + {(char *) "virDomainGetBlockIoTune", libvirt_virDomainGetBlockIoTune, METH_VARARGS, NULL}, {(char *) "virDomainSendKey", libvirt_virDomainSendKey, METH_VARARGS, NULL}, {(char *) "virDomainMigrateGetMaxSpeed", libvirt_virDomainMigrateGetMaxSpeed, METH_VARARGS, NULL}, {NULL, NULL, 0, NULL} -- 1.7.1

2 1

[libvirt] Wiki spam!
by Cole Robinson 10 Nov '11

10 Nov '11

The wiki's been spammed, see November 9 for example: http://wiki.libvirt.org/page/Special:Recentchanges virt-manager.org had a similar problem. Any one know any good proactive ways to prevent this? - Cole

2 3

[libvirt] [PATCH 5/8] Enable the blkdeviotune command in virsh
by Lei Li 10 Nov '11

10 Nov '11

Support virsh command blkdeviotune. Can set or query a block disk I/O throttle setting. Signed-off-by: Zhi Yong Wu <wuzhy(a)linux.vnet.ibm.com> Signed-off-by: Lei Li <lilei(a)linux.vnet.ibm.com> --- tools/virsh.c | 146 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ tools/virsh.pod | 23 +++++++++ 2 files changed, 169 insertions(+), 0 deletions(-) diff --git a/tools/virsh.c b/tools/virsh.c index 83dc3c7..df2b399 100644 --- a/tools/virsh.c +++ b/tools/virsh.c @@ -6023,6 +6023,151 @@ cmdBlockJob(vshControl *ctl, const vshCmd *cmd) return true; } +/* + * "blkdeviotune" command + */ +static const vshCmdInfo info_blkdeviotune[] = { + {"help", N_("Set or query a block disk I/O throttle setting.")}, + {"desc", N_("Set or query a block disk I/O throttle setting.\n" \ + " To query the block disk I/O throttle setting use the following" \ + " command: \n\n" \ + " virsh # blkdeviotune <domain> <device>")}, + {NULL, NULL} +}; + +static const vshCmdOptDef opts_blkdeviotune[] = { + {"domain", VSH_OT_DATA, VSH_OFLAG_REQ, N_("domain name, id or uuid")}, + {"device", VSH_OT_DATA, VSH_OFLAG_REQ, N_("block device")}, + {"total_bytes_sec", VSH_OT_INT, VSH_OFLAG_NONE, N_("total throughput limit in bytes per second/s")}, + {"read_bytes_sec", VSH_OT_INT, VSH_OFLAG_NONE, N_("read throughput limit in bytes per second/s")}, + {"write_bytes_sec", VSH_OT_INT, VSH_OFLAG_NONE, N_("write throughput limit in bytes per second/s")}, + {"total_iops_sec", VSH_OT_INT, VSH_OFLAG_NONE, N_("total I/O operations limit per second/s")}, + {"read_iops_sec", VSH_OT_INT, VSH_OFLAG_NONE, N_("read I/O operations limit per second/s")}, + {"write_iops_sec", VSH_OT_INT, VSH_OFLAG_NONE, N_("write I/O operations limit per second/s")}, + {"config", VSH_OT_BOOL, 0, N_("affect next boot")}, + {"live", VSH_OT_BOOL, 0, N_("affect running domain")}, + {"current", VSH_OT_BOOL, 0, N_("affect current domain")}, + {NULL, 0, 0, NULL} +}; + +static bool +cmdBlkdeviotune(vshControl *ctl, const vshCmd *cmd) +{ + virDomainPtr dom = NULL; + const char *name, *disk; + virDomainBlockIoTuneInfo info; + virDomainBlockIoTuneInfo reply; + unsigned int flags = 0; + int ret = -1; + int rv = 0; + int set = 0; + int current = vshCommandOptBool(cmd, "current"); + int config = vshCommandOptBool(cmd, "config"); + int live = vshCommandOptBool(cmd, "live"); + + if (current) { + if (live || config) { + vshError(ctl, "%s", _("--current must be specified exclusively")); + return false; + } + flags = VIR_DOMAIN_AFFECT_CURRENT; + } else { + if (config) + flags |= VIR_DOMAIN_AFFECT_CONFIG; + if (live) + flags |= VIR_DOMAIN_AFFECT_LIVE; + } + + memset(&info, 0, sizeof(info)); + + if (!vshConnectionUsability(ctl, ctl->conn)) + goto out; + + if (!(dom = vshCommandOptDomain(ctl, cmd, &name))) + goto out; + + if (vshCommandOptString(cmd, "device", &disk) < 0) + goto out; + + if ((rv = vshCommandOptULongLong(cmd, "total_bytes_sec", &info.total_bytes_sec)) < 0) { + vshError(ctl, "%s", + _("Unable to parse integer parameter")); + goto out; + } else if (rv > 0) { + set++; + } + + if ((rv = vshCommandOptULongLong(cmd, "read_bytes_sec", &info.read_bytes_sec)) < 0) { + vshError(ctl, "%s", + _("Unable to parse integer parameter")); + goto out; + } else if (rv > 0) { + set++; + } + + if ((rv = vshCommandOptULongLong(cmd, "write_bytes_sec", &info.write_bytes_sec)) < 0) { + vshError(ctl, "%s", + _("Unable to parse integer parameter")); + goto out; + } else if (rv > 0) { + set++; + } + + if ((rv = vshCommandOptULongLong(cmd, "total_iops_sec", &info.total_iops_sec)) < 0) { + vshError(ctl, "%s", + _("Unable to parse integer parameter")); + goto out; + } else if (rv > 0) { + set++; + } + + if ((rv = vshCommandOptULongLong(cmd, "read_iops_sec", &info.read_iops_sec)) < 0) { + vshError(ctl, "%s", + _("Unable to parse integer parameter")); + goto out; + } else if (rv > 0) { + set++; + } + + if ((rv = vshCommandOptULongLong(cmd, "write_iops_sec", &info.write_iops_sec)) < 0) { + vshError(ctl, "%s", + _("Unable to parse integer parameter")); + goto out; + } else if (rv > 0) { + set++; + } + + if (!set) { + + ret = virDomainGetBlockIoTune(dom, disk, &reply, flags); + + if (ret != 0) + goto out; + + vshPrint(ctl, "%-15s %llu\n", _("total_bytes_sec:"), reply.total_bytes_sec); + vshPrint(ctl, "%-15s %llu\n", _("read_bytes_sec:"), reply.read_bytes_sec); + vshPrint(ctl, "%-15s %llu\n", _("write_bytes_sec:"), reply.write_bytes_sec); + vshPrint(ctl, "%-15s %llu\n", _("total_iops_sec:"), reply.total_iops_sec); + vshPrint(ctl, "%-15s %llu\n", _("read_iops_sec:"), reply.read_iops_sec); + vshPrint(ctl, "%-15s %llu\n", _("write_iops_sec:"), reply.write_iops_sec); + + virDomainFree(dom); + return true; + } else { + + ret = virDomainSetBlockIoTune(dom, disk, &info, flags); + + if (ret == 0) { + virDomainFree(dom); + return true; + } + } + +out: + virDomainFree(dom); + return false; +} + /* * "net-autostart" command @@ -14023,6 +14168,7 @@ static const vshCmdDef domManagementCmds[] = { {"blkiotune", cmdBlkiotune, opts_blkiotune, info_blkiotune, 0}, {"blockpull", cmdBlockPull, opts_block_pull, info_block_pull, 0}, {"blockjob", cmdBlockJob, opts_block_job, info_block_job, 0}, + {"blkdeviotune", cmdBlkdeviotune, opts_blkdeviotune, info_blkdeviotune, 0}, #ifndef WIN32 {"console", cmdConsole, opts_console, info_console, 0}, #endif diff --git a/tools/virsh.pod b/tools/virsh.pod index 775d302..58fcb51 100644 --- a/tools/virsh.pod +++ b/tools/virsh.pod @@ -572,6 +572,29 @@ operation can be checked with B<blockjob>. I<path> specifies fully-qualified path of the disk. I<bandwidth> specifies copying bandwidth limit in Mbps. +=item B<blkdeviotune> I<domain> I<device> [[I<--total_bytes_sec> B<total_bytes_sec>] +| [[I<--read_bytes_sec> B<read_bytes_sec>] [I<--write_bytes_sec> B<write_bytes_sec>]] +[[I<--total_iops_sec> B<total_iops_sec>] | [[I<--read_iops_sec> B<read_iops_sec>] +[I<--write_iops_sec> B<write_iops_sec>]] [[I<--config>] [I<--live>] | [I<--current>]] + +Set or query the block disk io limits settting. +I<path> specifies block disk name. +I<--total_bytes_sec> specifies total throughput limit in bytes per second/s. +I<--read_bytes_sec> specifies read throughput limit in bytes per second/s. +I<--write_bytes_sec> specifies write throughput limit in bytes per second/s. +I<--total_iops_sec> specifies total I/O operations limit per second/s. +I<--read_iops_sec> specifies read I/O operations limit per second/s. +I<--write_iops_sec> specifies write I/O operations limit per second/s. + +If I<--live> is specified, affect a running guest. +If I<--config> is specified, affect the next boot of a persistent guest. +If I<--current> is specified, affect the current guest state. +Both I<--live> and I<--current> flags may be given, but I<--current> is +exclusive. If no flag is specified, behavior is different depending +on hypervisor. + +If no limit is specified, it will query current I/O limits setting. + =item B<blockjob> I<domain> I<path> [I<--abort>] [I<--info>] [I<bandwidth>] Manage active block operations. -- 1.7.1

2 1

[libvirt] [PATCH 4/8] Support block I/O throtte in XML
by Lei Li 10 Nov '11

10 Nov '11

Enable block I/O throttle for per-disk in XML. Signed-off-by: Zhi Yong Wu <wuzhy(a)linux.vnet.ibm.com> Signed-off-by: Lei Li <lilei(a)linux.vnet.ibm.com> --- src/conf/domain_conf.c | 101 +++++++++++++++++++++++++++++++++++++++++++++- src/conf/domain_conf.h | 12 ++++++ src/qemu/qemu_command.c | 33 +++++++++++++++ src/util/xml.h | 2 + 4 files changed, 145 insertions(+), 3 deletions(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 58f4d0f..564914e 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -2318,7 +2318,8 @@ static virDomainDiskDefPtr virDomainDiskDefParseXML(virCapsPtr caps, xmlNodePtr node, virBitmapPtr bootMap, - unsigned int flags) + unsigned int flags, + xmlXPathContextPtr ctxt) { virDomainDiskDefPtr def; xmlNodePtr cur, child; @@ -2517,6 +2518,62 @@ virDomainDiskDefParseXML(virCapsPtr caps, } child = child->next; } + } else if (xmlStrEqual(cur->name, BAD_CAST "iotune")) { + if (virXPathULongLong("string(./devices/disk/iotune/total_bytes_sec)", + ctxt, &def->blkdeviotune.total_bytes_sec) < 0) { + def->blkdeviotune.total_bytes_sec = 0; + } else { + def->blkdeviotune.mark = 1; + } + + if (virXPathULongLong("string(./devices/disk/iotune/read_bytes_sec)", + ctxt, &def->blkdeviotune.read_bytes_sec) < 0) { + def->blkdeviotune.read_bytes_sec = 0; + } else { + def->blkdeviotune.mark = 1; + } + + if (virXPathULongLong("string(./devices/disk/iotune/write_bytes_sec)", + ctxt, &def->blkdeviotune.write_bytes_sec) < 0) { + def->blkdeviotune.write_bytes_sec = 0; + } else { + def->blkdeviotune.mark = 1; + } + + if (virXPathULongLong("string(./devices/disk/iotune/total_iops_sec)", + ctxt, &def->blkdeviotune.total_iops_sec) < 0) { + def->blkdeviotune.total_iops_sec = 0; + } else { + def->blkdeviotune.mark = 1; + } + + if (virXPathULongLong("string(./devices/disk/iotune/read_iops_sec)", + ctxt, &def->blkdeviotune.read_iops_sec) < 0) { + def->blkdeviotune.read_iops_sec = 0; + } else { + def->blkdeviotune.mark = 1; + } + + if (virXPathULongLong("string(./devices/disk/iotune/write_iops_sec)", + ctxt, &def->blkdeviotune.write_iops_sec) < 0) { + def->blkdeviotune.write_iops_sec = 0; + } else { + def->blkdeviotune.mark = 1; + } + + if ((def->blkdeviotune.total_bytes_sec && def->blkdeviotune.read_bytes_sec) + || (def->blkdeviotune.total_bytes_sec && def->blkdeviotune.write_bytes_sec)) { + virDomainReportError(VIR_ERR_XML_ERROR, + _("bps and bps_rd/bps_wr cannot be set at the same time")); + goto error; + } + + if ((def->blkdeviotune.total_iops_sec && def->blkdeviotune.read_iops_sec) + || (def->blkdeviotune.total_iops_sec && def->blkdeviotune.write_iops_sec)) { + virDomainReportError(VIR_ERR_XML_ERROR, + _("iops and iops_rd/iops_wr cannot be set at the same time")); + goto error; + } } else if (xmlStrEqual(cur->name, BAD_CAST "readonly")) { def->readonly = 1; } else if (xmlStrEqual(cur->name, BAD_CAST "shareable")) { @@ -6003,7 +6060,7 @@ virDomainDeviceDefPtr virDomainDeviceDefParse(virCapsPtr caps, if (xmlStrEqual(node->name, BAD_CAST "disk")) { dev->type = VIR_DOMAIN_DEVICE_DISK; if (!(dev->data.disk = virDomainDiskDefParseXML(caps, node, - NULL, flags))) + NULL, flags, NULL))) goto error; } else if (xmlStrEqual(node->name, BAD_CAST "lease")) { dev->type = VIR_DOMAIN_DEVICE_LEASE; @@ -7076,7 +7133,8 @@ static virDomainDefPtr virDomainDefParseXML(virCapsPtr caps, virDomainDiskDefPtr disk = virDomainDiskDefParseXML(caps, nodes[i], bootMap, - flags); + flags, + ctxt); if (!disk) goto error; @@ -9511,6 +9569,43 @@ virDomainDiskDefFormat(virBufferPtr buf, virBufferAsprintf(buf, " <target dev='%s' bus='%s'/>\n", def->dst, bus); + /*disk I/O throttling*/ + if (def->blkdeviotune.mark) { + virBufferAddLit(buf, " <iotune>\n"); + if (def->blkdeviotune.total_bytes_sec) { + virBufferAsprintf(buf, " <total_bytes_sec>%llu</total_bytes_sec>\n", + def->blkdeviotune.total_bytes_sec); + } + + if (def->blkdeviotune.read_bytes_sec) { + virBufferAsprintf(buf, " <read_bytes_sec>%llu</read_bytes_sec>\n", + def->blkdeviotune.read_bytes_sec); + + } + + if (def->blkdeviotune.write_bytes_sec) { + virBufferAsprintf(buf, " <write_bytes_sec>%llu</write_bytes_sec>\n", + def->blkdeviotune.write_bytes_sec); + } + + if (def->blkdeviotune.total_iops_sec) { + virBufferAsprintf(buf, " <total_iops_sec>%llu</total_iops_sec>\n", + def->blkdeviotune.total_iops_sec); + } + + if (def->blkdeviotune.read_iops_sec) { + virBufferAsprintf(buf, " <read_iops_sec>%llu</read_iops_sec>", + def->blkdeviotune.read_iops_sec); + } + + if (def->blkdeviotune.write_iops_sec) { + virBufferAsprintf(buf, " <write_iops_sec>%llu</write_iops_sec>", + def->blkdeviotune.write_iops_sec); + } + + virBufferAddLit(buf, " </iotune>\n"); + } + if (def->bootIndex) virBufferAsprintf(buf, " <boot order='%d'/>\n", def->bootIndex); if (def->readonly) diff --git a/src/conf/domain_conf.h b/src/conf/domain_conf.h index a3cb834..d95e239 100644 --- a/src/conf/domain_conf.h +++ b/src/conf/domain_conf.h @@ -333,6 +333,18 @@ struct _virDomainDiskDef { } auth; char *driverName; char *driverType; + + /*disk I/O throttling*/ + struct { + unsigned long long total_bytes_sec; + unsigned long long read_bytes_sec; + unsigned long long write_bytes_sec; + unsigned long long total_iops_sec; + unsigned long long read_iops_sec; + unsigned long long write_iops_sec; + unsigned int mark; + } blkdeviotune; + char *serial; int cachemode; int error_policy; /* enum virDomainDiskErrorPolicy */ diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index 11ebb69..3a8575b 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -1715,6 +1715,39 @@ qemuBuildDriveStr(virConnectPtr conn ATTRIBUTE_UNUSED, } } + /*block I/O throttling*/ + if (disk->blkdeviotune.mark) { + if (disk->blkdeviotune.total_bytes_sec) { + virBufferAsprintf(&opt, ",bps=%llu", + disk->blkdeviotune.total_bytes_sec); + } + + if (disk->blkdeviotune.read_bytes_sec) { + virBufferAsprintf(&opt, ",bps_rd=%llu", + disk->blkdeviotune.read_bytes_sec); + } + + if (disk->blkdeviotune.write_bytes_sec) { + virBufferAsprintf(&opt, ",bps_wr=%llu", + disk->blkdeviotune.write_bytes_sec); + } + + if (disk->blkdeviotune.total_iops_sec) { + virBufferAsprintf(&opt, ",iops=%llu", + disk->blkdeviotune.total_iops_sec); + } + + if (disk->blkdeviotune.read_iops_sec) { + virBufferAsprintf(&opt, ",iops_rd=%llu", + disk->blkdeviotune.read_iops_sec); + } + + if (disk->blkdeviotune.write_iops_sec) { + virBufferAsprintf(&opt, ",iops_wr=%llu", + disk->blkdeviotune.write_iops_sec); + } + } + if (virBufferError(&opt)) { virReportOOMError(); goto error; diff --git a/src/util/xml.h b/src/util/xml.h index c492063..5742f19 100644 --- a/src/util/xml.h +++ b/src/util/xml.h @@ -50,6 +50,8 @@ xmlNodePtr virXPathNode(const char *xpath, int virXPathNodeSet(const char *xpath, xmlXPathContextPtr ctxt, xmlNodePtr **list); +int virXMLStringToULongLong (const char *stringval, + unsigned long long *value); char * virXMLPropString(xmlNodePtr node, const char *name); -- 1.7.1

2 1

[libvirt] [PATCH 3/8] Implement virDomain{Set, Get}BlockIoTune for the qemu driver
by Lei Li 10 Nov '11

10 Nov '11

This patch implement the blk io throttle setting and getting support to qemu driver. Signed-off-by: Zhi Yong Wu <wuzhy(a)linux.vnet.ibm.com> Signed-off-by: Lei Li <lilei(a)linux.vnet.ibm.com> --- src/qemu/qemu_driver.c | 216 ++++++++++++++++++++++++++++++++++++++++++ src/qemu/qemu_monitor.c | 36 +++++++ src/qemu/qemu_monitor.h | 10 ++ src/qemu/qemu_monitor_json.c | 191 +++++++++++++++++++++++++++++++++++++ src/qemu/qemu_monitor_json.h | 10 ++ src/qemu/qemu_monitor_text.c | 152 +++++++++++++++++++++++++++++ src/qemu/qemu_monitor_text.h | 10 ++ 7 files changed, 625 insertions(+), 0 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index db2ac0d..7ca6719 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -10759,6 +10759,220 @@ cleanup: return ret; } +static int +qemuDomainSetBlockIoTune(virDomainPtr dom, + const char *disk, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags) +{ + struct qemud_driver *driver = dom->conn->privateData; + virDomainObjPtr vm = NULL; + qemuDomainObjPrivatePtr priv; + virDomainDefPtr persistentDef = NULL; + char uuidstr[VIR_UUID_STRING_BUFLEN]; + const char *device = NULL; + int ret = -1; + bool isActive; + + virCheckFlags(VIR_DOMAIN_AFFECT_LIVE | + VIR_DOMAIN_AFFECT_CONFIG, -1); + + qemuDriverLock(driver); + virUUIDFormat(dom->uuid, uuidstr); + vm = virDomainFindByUUID(&driver->domains, dom->uuid); + if (!vm) { + qemuReportError(VIR_ERR_NO_DOMAIN, + _("no domain with matching uuid '%s'"), uuidstr); + goto cleanup; + } + + device = qemuDiskPathToAlias(vm, disk); + if (!device) { + goto cleanup; + } + + if (qemuDomainObjBeginJobWithDriver(driver, vm, QEMU_JOB_MODIFY) < 0) + goto cleanup; + + isActive = virDomainObjIsActive(vm); + + if (flags == VIR_DOMAIN_AFFECT_CURRENT) { + if (isActive) + flags = VIR_DOMAIN_AFFECT_LIVE; + else + flags = VIR_DOMAIN_AFFECT_CONFIG; + } + + if ((info->total_bytes_sec && info->read_bytes_sec) || + (info->total_bytes_sec && info->write_bytes_sec)) { + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s", + _("bps and bps_rd/bps_wr cannot be used at the same time")); + goto endjob; + } + + if ((info->total_iops_sec && info->read_iops_sec) || + (info->total_iops_sec && info->write_iops_sec)) { + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s", + _("iops and iops_rd/iops_wr cannot be used at the same time")); + goto endjob; + } + + if (!isActive && (flags & VIR_DOMAIN_AFFECT_LIVE)) { + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s", + _("domain is not running")); + goto endjob; + } + + if (flags & VIR_DOMAIN_AFFECT_CONFIG) { + if (!vm->persistent) { + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s", + _("cannot change persistent config of a transient domain")); + goto endjob; + } + if (!(persistentDef = virDomainObjGetPersistentDef(driver->caps, vm))) + goto endjob; + } + + if (flags & VIR_DOMAIN_AFFECT_LIVE) { + priv = vm->privateData; + qemuDomainObjEnterMonitorWithDriver(driver, vm); + ret = qemuMonitorSetBlockIoThrottle(priv->mon, device, info, 1); + qemuDomainObjExitMonitorWithDriver(driver, vm); + } + + if (flags & VIR_DOMAIN_AFFECT_CONFIG) { + sa_assert(persistentDef); + int i = virDomainDiskIndexByName(vm->def, disk, true); + if (i < 0) + goto endjob; + persistentDef->disks[i]->blkdeviotune.total_bytes_sec = info->total_bytes_sec; + persistentDef->disks[i]->blkdeviotune.read_bytes_sec = info->read_bytes_sec; + persistentDef->disks[i]->blkdeviotune.write_bytes_sec = info->write_bytes_sec; + persistentDef->disks[i]->blkdeviotune.total_iops_sec = info->total_iops_sec; + persistentDef->disks[i]->blkdeviotune.read_iops_sec = info->read_iops_sec; + persistentDef->disks[i]->blkdeviotune.write_iops_sec = info->write_iops_sec; + persistentDef->disks[i]->blkdeviotune.mark = 1; + } + + if (flags & VIR_DOMAIN_AFFECT_CONFIG) { + ret = virDomainSaveConfig(driver->configDir, persistentDef); + if (ret < 0) { + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s", + _("Write to config file failed")); + goto endjob; + } + } + +endjob: + if (qemuDomainObjEndJob(driver, vm) == 0) + vm = NULL; + +cleanup: + VIR_FREE(device); + if (vm) + virDomainObjUnlock(vm); + qemuDriverUnlock(driver); + return ret; +} + +static int +qemuDomainGetBlockIoTune(virDomainPtr dom, + const char *disk, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags) +{ + struct qemud_driver *driver = dom->conn->privateData; + virDomainObjPtr vm = NULL; + qemuDomainObjPrivatePtr priv; + virDomainDefPtr persistentDef = NULL; + char uuidstr[VIR_UUID_STRING_BUFLEN]; + const char *device = NULL; + int ret = -1; + bool isActive; + + virCheckFlags(VIR_DOMAIN_AFFECT_LIVE | + VIR_DOMAIN_AFFECT_CONFIG, -1); + + qemuDriverLock(driver); + virUUIDFormat(dom->uuid, uuidstr); + vm = virDomainFindByUUID(&driver->domains, dom->uuid); + if (!vm) { + qemuReportError(VIR_ERR_NO_DOMAIN, + _("no domain with matching uuid '%s'"), uuidstr); + goto cleanup; + } + + device = qemuDiskPathToAlias(vm, disk); + + if (!device) { + goto cleanup; + } + + if (qemuDomainObjBeginJobWithDriver(driver, vm, QEMU_JOB_MODIFY) < 0) + goto cleanup; + + isActive = virDomainObjIsActive(vm); + + if (flags == VIR_DOMAIN_AFFECT_CURRENT) { + if (isActive) + flags = VIR_DOMAIN_AFFECT_LIVE; + else + flags = VIR_DOMAIN_AFFECT_CONFIG; + } + + if ((flags & VIR_DOMAIN_AFFECT_LIVE) && (flags & VIR_DOMAIN_AFFECT_CONFIG)) { + qemuReportError(VIR_ERR_INVALID_ARG, "%s", + _("Cannot query with --live and --config together")); + goto cleanup; + } + + if (!isActive && (flags & VIR_DOMAIN_AFFECT_LIVE)) { + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s", + _("domain is not running")); + goto cleanup; + } + + if (flags & VIR_DOMAIN_AFFECT_LIVE) { + priv = vm->privateData; + qemuDomainObjEnterMonitorWithDriver(driver, vm); + ret = qemuMonitorGetBlockIoThrottle(priv->mon, device, reply, 0); + qemuDomainObjExitMonitorWithDriver(driver, vm); + } + + if (flags & VIR_DOMAIN_AFFECT_CONFIG) { + if (!vm->persistent) { + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s", + _("domain is transient")); + goto cleanup; + } + if (!(persistentDef = virDomainObjGetPersistentDef(driver->caps, vm))) + goto cleanup; + + sa_assert(persistentDef); + int i = virDomainDiskIndexByName(vm->def, disk, true); + if (i < 0) + goto cleanup; + reply->total_bytes_sec = persistentDef->disks[i]->blkdeviotune.total_bytes_sec; + reply->read_bytes_sec = persistentDef->disks[i]->blkdeviotune.read_bytes_sec; + reply->write_bytes_sec = persistentDef->disks[i]->blkdeviotune.write_bytes_sec; + reply->total_iops_sec = persistentDef->disks[i]->blkdeviotune.total_iops_sec; + reply->read_iops_sec = persistentDef->disks[i]->blkdeviotune.read_iops_sec; + reply->write_iops_sec = persistentDef->disks[i]->blkdeviotune.write_iops_sec; + ret = 0; + } + + if (qemuDomainObjEndJob(driver, vm) == 0) { + vm = NULL; + goto cleanup; + } + +cleanup: + VIR_FREE(device); + if (vm) + virDomainObjUnlock(vm); + qemuDriverUnlock(driver); + return ret; +} static virDriver qemuDriver = { .no = VIR_DRV_QEMU, @@ -10903,6 +11117,8 @@ static virDriver qemuDriver = { .domainGetBlockJobInfo = qemuDomainGetBlockJobInfo, /* 0.9.4 */ .domainBlockJobSetSpeed = qemuDomainBlockJobSetSpeed, /* 0.9.4 */ .domainBlockPull = qemuDomainBlockPull, /* 0.9.4 */ + .domainSetBlockIoTune = qemuDomainSetBlockIoTune, /* 0.9.8 */ + .domainGetBlockIoTune = qemuDomainGetBlockIoTune, /* 0.9.8 */ }; diff --git a/src/qemu/qemu_monitor.c b/src/qemu/qemu_monitor.c index 73e5ea9..5f5bd08 100644 --- a/src/qemu/qemu_monitor.c +++ b/src/qemu/qemu_monitor.c @@ -2567,6 +2567,42 @@ int qemuMonitorBlockJob(qemuMonitorPtr mon, return ret; } +int qemuMonitorSetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags) +{ + int ret; + + VIR_DEBUG("mon=%p, device=%p, info=%p, flags=%x", + mon, device, info, flags); + + if (mon->json) { + ret = qemuMonitorJSONSetBlockIoThrottle(mon, device, info, flags); + } else { + ret = qemuMonitorTextSetBlockIoThrottle(mon, device, info, flags); + } + return ret; +} + +int qemuMonitorGetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags) +{ + int ret; + + VIR_DEBUG("mon=%p, device=%p, reply=%p, flags=%x", + mon, device, reply, flags); + + if (mon->json) { + ret = qemuMonitorJSONGetBlockIoThrottle(mon, device, reply, flags); + } else { + ret = qemuMonitorTextGetBlockIoThrottle(mon, device, reply, flags); + } + return ret; +} + int qemuMonitorVMStatusToPausedReason(const char *status) { int st; diff --git a/src/qemu/qemu_monitor.h b/src/qemu/qemu_monitor.h index 883e0aa..bf00068 100644 --- a/src/qemu/qemu_monitor.h +++ b/src/qemu/qemu_monitor.h @@ -521,6 +521,16 @@ int qemuMonitorOpenGraphics(qemuMonitorPtr mon, const char *fdname, bool skipauth); +int qemuMonitorSetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags); + +int qemuMonitorGetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags); + /** * When running two dd process and using <> redirection, we need a * shell that will not truncate files. These two strings serve that diff --git a/src/qemu/qemu_monitor_json.c b/src/qemu/qemu_monitor_json.c index 56a62db..7501c59 100644 --- a/src/qemu/qemu_monitor_json.c +++ b/src/qemu/qemu_monitor_json.c @@ -3276,3 +3276,194 @@ int qemuMonitorJSONOpenGraphics(qemuMonitorPtr mon, virJSONValueFree(reply); return ret; } + +static int +qemuMonitorJSONBlockIoThrottleInfo(virJSONValuePtr result, + const char *device, + virDomainBlockIoTuneInfoPtr reply) +{ + virJSONValuePtr io_throttle; + int ret = -1; + int i; + int found = 0; + + io_throttle = virJSONValueObjectGet(result, "return"); + + if (!io_throttle || io_throttle->type != VIR_JSON_TYPE_ARRAY) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _(" block_io_throttle reply was missing device list ")); + goto cleanup; + } + + for (i = 0; i < virJSONValueArraySize(io_throttle); i++) { + virJSONValuePtr temp_dev = virJSONValueArrayGet(io_throttle, i); + virJSONValuePtr inserted; + const char *current_dev; + + if (!temp_dev || temp_dev->type != VIR_JSON_TYPE_OBJECT) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("block io throttle device entry was not in expected format")); + goto cleanup; + } + + if ((current_dev = virJSONValueObjectGetString(temp_dev, "device")) == NULL) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("block io throttle device entry was not in expected format")); + goto cleanup; + } + + if(STRPREFIX(current_dev, QEMU_DRIVE_HOST_PREFIX)) + current_dev += strlen(QEMU_DRIVE_HOST_PREFIX); + + if (STREQ(current_dev, device)) + continue; + + found = 1; + if ((inserted = virJSONValueObjectGet(temp_dev, "inserted")) == NULL || + inserted->type != VIR_JSON_TYPE_OBJECT) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("block io throttle inserted entry was not in expected format")); + goto cleanup; + } + + if (virJSONValueObjectGetNumberUlong(inserted, "bps", &reply->total_bytes_sec) < 0) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + _("cannot read %s"), + "total_bytes_sec"); + goto cleanup; + } + + if (virJSONValueObjectGetNumberUlong(inserted, "bps_rd", &reply->read_bytes_sec) < 0) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + _("cannot read %s"), + "read_bytes_sec"); + goto cleanup; + } + + if (virJSONValueObjectGetNumberUlong(inserted, "bps_wr", &reply->write_bytes_sec) < 0) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + _("cannot read %s"), + "write_bytes_sec"); + goto cleanup; + } + + if (virJSONValueObjectGetNumberUlong(inserted, "iops", &reply->total_iops_sec) < 0) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + _("cannot read %s"), + "total_iops_sec"); + goto cleanup; + } + + if (virJSONValueObjectGetNumberUlong(inserted, "iops_rd", &reply->read_iops_sec) < 0) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + _("cannot read %s"), + "read_iops_sec"); + goto cleanup; + } + + if (virJSONValueObjectGetNumberUlong(inserted, "iops_wr", &reply->write_iops_sec) < 0) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + _("cannot read %s"), + "write_iops_sec"); + goto cleanup; + } + break; + } + + if (!found) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + _("cannot found info for device '%s'"), + device); + goto cleanup; + } + ret = 0; + +cleanup: + return ret; +} + +int qemuMonitorJSONSetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags) +{ + int ret = -1; + virJSONValuePtr cmd = NULL; + virJSONValuePtr result = NULL; + + if (flags) { + cmd = qemuMonitorJSONMakeCommand("block_set_io_throttle", + "s:device", device, + "U:bps", info->total_bytes_sec, + "U:bps_rd", info->read_bytes_sec, + "U:bps_wr", info->write_bytes_sec, + "U:iops", info->total_iops_sec, + "U:iops_rd", info->read_iops_sec, + "U:iops_wr", info->write_iops_sec, + NULL); + } + + if (!cmd) + return -1; + + ret = qemuMonitorJSONCommand(mon, cmd, &result); + + if (ret == 0 && virJSONValueObjectHasKey(result, "error")) { + if (qemuMonitorJSONHasError(result, "DeviceNotActive")) + qemuReportError(VIR_ERR_OPERATION_INVALID, + _("No active operation on device: %s"), device); + else if (qemuMonitorJSONHasError(result, "NotSupported")) + qemuReportError(VIR_ERR_OPERATION_INVALID, + _("Operation is not supported for device: %s"), device); + else + qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("Unexpected error")); + ret = -1; + } + + virJSONValueFree(cmd); + virJSONValueFree(result); + return ret; +} + +int qemuMonitorJSONGetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags) +{ + int ret = -1; + virJSONValuePtr cmd = NULL; + virJSONValuePtr result = NULL; + + if (!flags) { + cmd = qemuMonitorJSONMakeCommand("query-block", + NULL); + } + + if (!cmd) { + return -1; + } + + ret = qemuMonitorJSONCommand(mon, cmd, &result); + + if (ret == 0 && virJSONValueObjectHasKey(result, "error")) { + if (qemuMonitorJSONHasError(result, "DeviceNotActive")) + qemuReportError(VIR_ERR_OPERATION_INVALID, + _("No active operation on device: %s"), device); + else if (qemuMonitorJSONHasError(result, "NotSupported")) + qemuReportError(VIR_ERR_OPERATION_INVALID, + _("Operation is not supported for device: %s"), device); + else + qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("Unexpected error")); + ret = -1; + } + + if (ret == 0 && !flags) + ret = qemuMonitorJSONBlockIoThrottleInfo(result, device, reply); + + virJSONValueFree(cmd); + virJSONValueFree(result); + return ret; +} + diff --git a/src/qemu/qemu_monitor_json.h b/src/qemu/qemu_monitor_json.h index f10d7d2..bf12dc5 100644 --- a/src/qemu/qemu_monitor_json.h +++ b/src/qemu/qemu_monitor_json.h @@ -255,4 +255,14 @@ int qemuMonitorJSONOpenGraphics(qemuMonitorPtr mon, const char *fdname, bool skipauth); +int qemuMonitorJSONSetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags); + +int qemuMonitorJSONGetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags); + #endif /* QEMU_MONITOR_JSON_H */ diff --git a/src/qemu/qemu_monitor_text.c b/src/qemu/qemu_monitor_text.c index 5de4d24..5c92e60 100644 --- a/src/qemu/qemu_monitor_text.c +++ b/src/qemu/qemu_monitor_text.c @@ -3429,3 +3429,155 @@ cleanup: VIR_FREE(cmd); return ret; } + +int qemuMonitorTextSetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags) +{ + char *cmd = NULL; + char *result = NULL; + int ret = 0; + + if (flags) { + ret = virAsprintf(&cmd, "block_set_io_throttle %s %llu %llu %llu %llu %llu %llu", + device, info->total_bytes_sec, info->read_bytes_sec, + info->write_bytes_sec, info->total_iops_sec, + info->read_iops_sec, info->write_iops_sec); + } + + if (ret < 0) { + virReportOOMError(); + return -1; + } + + if (qemuMonitorHMPCommand(mon, cmd, &result) < 0) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + "%s", _("cannot run monitor command")); + ret = -1; + goto cleanup; + } + +cleanup: + VIR_FREE(cmd); + VIR_FREE(result); + return ret; +} + +static int qemuMonitorTextParseBlockIoThrottle(const char *result, + const char *device, + virDomainBlockIoTuneInfoPtr reply) +{ + char *dummy = NULL; + int ret = -1; + const char *p, *eol; + int devnamelen = strlen(device); + + p = result; + + while (*p) { + if (STRPREFIX(p, QEMU_DRIVE_HOST_PREFIX)) + p += strlen(QEMU_DRIVE_HOST_PREFIX); + + if (STREQLEN(p, device, devnamelen) && + p[devnamelen] == ':' && p[devnamelen+1] == ' ') { + + eol = strchr(p, '\n'); + if (!eol) + eol = p + strlen(p); + + p += devnamelen + 2; /*Skip to first label. */ + + while (*p) { + if (STRPREFIX(p, "bps=")) { + p += strlen("bps="); + if (virStrToLong_ull(p, &dummy, 10, &reply->total_bytes_sec) == -1) + VIR_DEBUG("error reading total_bytes_sec: %s", p); + } else if (STRPREFIX(p, "bps_rd=")) { + p += strlen("bps_rd="); + if (virStrToLong_ull(p, &dummy, 10, &reply->read_bytes_sec) == -1) + VIR_DEBUG("error reading read_bytes_sec: %s", p); + } else if (STRPREFIX(p, "bps_wr=")) { + p += strlen("bps_wr="); + if (virStrToLong_ull(p, &dummy, 10, &reply->write_bytes_sec) == -1) + VIR_DEBUG("error reading write_bytes_sec: %s", p); + } else if (STRPREFIX(p, "iops=")) { + p += strlen("iops="); + if (virStrToLong_ull(p, &dummy, 10, &reply->total_iops_sec) == -1) + VIR_DEBUG("error reading total_iops_sec: %s", p); + } else if (STRPREFIX(p, "iops_rd=")) { + p += strlen("iops_rd="); + if (virStrToLong_ull(p, &dummy, 10, &reply->read_iops_sec) == -1) + VIR_DEBUG("error reading read_iops_sec: %s", p); + } else if (STRPREFIX(p, "iops_wr=")) { + p += strlen("iops_wr="); + if (virStrToLong_ull(p, &dummy, 10, &reply->write_iops_sec) == -1) + VIR_DEBUG("error reading write_iops_sec: %s", p); + } else { + VIR_DEBUG(" unknown block info %s", p); + } + + /* Skip to next label. */ + p = strchr (p, ' '); + if (!p || p >= eol) + break; + p++; + } + ret = 0; + goto cleanup; + } + + /* Skip to next line. */ + p = strchr (p, '\n'); + if (!p) + break; + p++; + } + + qemuReportError(VIR_ERR_INVALID_ARG, + _("no info for device '%s'"), device); + +cleanup: + return ret; +} + +int qemuMonitorTextGetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags) +{ + char *cmd = NULL; + char *result = NULL; + int ret = 0; + + if (flags) { + ret = virAsprintf(&cmd, "info block"); + } + + if (ret < 0) { + virReportOOMError(); + return -1; + } + + if (qemuMonitorHMPCommand(mon, cmd, &result) < 0) { + qemuReportError(VIR_ERR_INTERNAL_ERROR, + "%s", _("cannot run monitor command")); + ret = -1; + goto cleanup; + } + + if (strstr(result, "\ninfo ")) { + qemuReportError(VIR_ERR_OPERATION_INVALID, + "%s", + _("info block not supported by this qemu")); + ret = -1; + goto cleanup; + } + ret = qemuMonitorTextParseBlockIoThrottle(result, device, reply); + +cleanup: + VIR_FREE(cmd); + VIR_FREE(result); + return ret; +} + diff --git a/src/qemu/qemu_monitor_text.h b/src/qemu/qemu_monitor_text.h index f32fce0..1c47d39 100644 --- a/src/qemu/qemu_monitor_text.h +++ b/src/qemu/qemu_monitor_text.h @@ -248,4 +248,14 @@ int qemuMonitorTextOpenGraphics(qemuMonitorPtr mon, const char *fdname, bool skipauth); +int qemuMonitorTextSetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags); + +int qemuMonitorTextGetBlockIoThrottle(qemuMonitorPtr mon, + const char *device, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags); + #endif /* QEMU_MONITOR_TEXT_H */ -- 1.7.1

2 1

[libvirt] Problem with libvirt daemon
by Wouter - Iamotor 10 Nov '11

10 Nov '11

Hello, I have a problem with the libvirt daemon. I have a little php script which should connect with libvirt, but it won't. When I try to start the libvirt daemon I get this error (using the command libvirtd --listen): 2011-11-09 19:35:18.095: 3560: info : libvirt version: 0.9.7 2011-11-09 19:35:18.095: 3560: error : virPidFileAcquirePath:345 : Failed to acquire pid file '/var/run/libvirtd.pid': Resource temporarily unavailable So the daemon won't start and isn't started yet. If I check if the service is running with ps -ef | grep libvirtd I get this output: root 2865 1 0 15:50 ? 00:00:00 /usr/sbin/libvirtd -d root 3149 1 0 19:17 ? 00:00:00 libvirtd -d root 3206 1 0 19:17 ? 00:00:00 libvirtd -dl root 3300 1 0 19:20 ? 00:00:00 libvirtd --listen --daemon root 3356 1 0 19:21 ? 00:00:00 libvirtd --listen --daemon root 3564 3077 0 19:36 pts/0 00:00:00 grep --color=auto libvirtd Here you can see that the service isn't running. How can I fix this problem. There is a .pid file in the /var/run/ directory. I already deleted it and used the command libvirtd -daemon -listen but it won't start. It doesn't show any output but when I use the ps -ef command, I see that the service is not running. Can you please help me? With kind regards, Wouter.

3 4

[libvirt] [PATCH 2/8] Add virDomain{Set, Get}BlockIoTune support to the remote driver
by Lei Li 10 Nov '11

10 Nov '11

Support Block I/O Throttle setting and query to remote driver. Signed-off-by: Zhi Yong Wu <wuzhy(a)linux.vnet.ibm.com> Signed-off-by: Lei Li <lilei(a)linux.vnet.ibm.com> --- daemon/remote.c | 87 ++++++++++++++++++++++++++++++++++++++++++ src/remote/remote_driver.c | 80 ++++++++++++++++++++++++++++++++++++++ src/remote/remote_protocol.x | 38 ++++++++++++++++++- src/remote_protocol-structs | 33 ++++++++++++++++ 4 files changed, 237 insertions(+), 1 deletions(-) diff --git a/daemon/remote.c b/daemon/remote.c index bd0c3e3..070ca65 100644 --- a/daemon/remote.c +++ b/daemon/remote.c @@ -1850,6 +1850,93 @@ cleanup: return rv; } +static int +remoteDispatchDomainSetBlockIoThrottle(virNetServerPtr server ATTRIBUTE_UNUSED, + virNetServerClientPtr client ATTRIBUTE_UNUSED, + virNetMessagePtr hdr ATTRIBUTE_UNUSED, + virNetMessageErrorPtr rerr, + remote_domain_set_block_io_throttle_args *args) +{ + virDomainPtr dom = NULL; + virDomainBlockIoTuneInfo tmp; + int rv = -1; + struct daemonClientPrivate *priv = + virNetServerClientGetPrivateData(client); + + if (!priv->conn) { + virNetError(VIR_ERR_INTERNAL_ERROR, "%s", _("connection not open")); + goto cleanup; + } + + if (!(dom = get_nonnull_domain(priv->conn, args->dom))) + goto cleanup; + + if (args) { + tmp.total_bytes_sec = args->bps; + tmp.read_bytes_sec = args->bps_rd; + tmp.write_bytes_sec = args->bps_wr; + tmp.total_iops_sec = args->iops; + tmp.read_iops_sec = args->iops_rd; + tmp.write_iops_sec = args->iops_wr; + } + + rv = virDomainSetBlockIoTune(dom, args->disk, &tmp, args->flags); + + if (rv < 0) + goto cleanup; + +cleanup: + if (rv < 0) + virNetMessageSaveError(rerr); + if (dom) + virDomainFree(dom); + return rv; +} + +static int +remoteDispatchDomainGetBlockIoThrottle(virNetServerPtr server ATTRIBUTE_UNUSED, + virNetServerClientPtr client ATTRIBUTE_UNUSED, + virNetMessagePtr hdr ATTRIBUTE_UNUSED, + virNetMessageErrorPtr rerr, + remote_domain_get_block_io_throttle_args *args, + remote_domain_get_block_io_throttle_ret *ret) +{ + virDomainPtr dom = NULL; + virDomainBlockIoTuneInfo reply; + int rv = -1; + struct daemonClientPrivate *priv = + virNetServerClientGetPrivateData(client); + + if (!priv->conn) { + virNetError(VIR_ERR_INTERNAL_ERROR, "%s", _("connection not open")); + goto cleanup; + } + + if (!(dom = get_nonnull_domain(priv->conn, args->dom))) + goto cleanup; + + rv = virDomainGetBlockIoTune(dom, args->disk, &reply, args->flags); + + if (rv < 0) { + ret->found = 0; + goto cleanup; + } + + ret->bps = reply.total_bytes_sec; + ret->bps_rd = reply.read_bytes_sec; + ret->bps_wr = reply.write_bytes_sec; + ret->iops = reply.total_iops_sec; + ret->iops_rd = reply.read_iops_sec; + ret->iops_wr = reply.write_iops_sec; + ret->found = 1; + +cleanup: + if (rv < 0) + virNetMessageSaveError(rerr); + if (dom) + virDomainFree(dom); + return rv; +} /*-------------------------------------------------------------*/ diff --git a/src/remote/remote_driver.c b/src/remote/remote_driver.c index f3b8ad5..0900231 100644 --- a/src/remote/remote_driver.c +++ b/src/remote/remote_driver.c @@ -2154,6 +2154,84 @@ done: return rv; } +static int remoteDomainSetBlockIoTune(virDomainPtr domain, + const char *disk, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags) +{ + int rv = -1; + remote_domain_set_block_io_throttle_args args; + struct private_data *priv = domain->conn->privateData; + + remoteDriverLock(priv); + + memset(&args, 0, sizeof(args)); + + make_nonnull_domain(&args.dom, domain); + args.disk = (char *)disk; + args.bps = info->total_bytes_sec; + args.bps_rd = info->read_bytes_sec; + args.bps_wr = info->write_bytes_sec; + args.iops = info->total_iops_sec; + args.iops_rd = info->read_iops_sec; + args.iops_wr = info->write_iops_sec; + args.flags = flags; + + if (call(domain->conn, priv, 0, REMOTE_PROC_DOMAIN_SET_BLOCK_IO_THROTTLE, + (xdrproc_t) xdr_remote_domain_set_block_io_throttle_args, + (char *) &args, + (xdrproc_t) xdr_void, + (char *) NULL) == -1) { + goto done; + } + rv = 0; + +done: + remoteDriverUnlock(priv); + return rv; +} + +static int remoteDomainGetBlockIoTune(virDomainPtr domain, + const char *disk, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags) +{ + int rv = -1; + remote_domain_get_block_io_throttle_args args; + remote_domain_get_block_io_throttle_ret ret; + struct private_data *priv = domain->conn->privateData; + + remoteDriverLock(priv); + + make_nonnull_domain(&args.dom, domain); + args.disk = (char *)disk; + args.flags = flags; + + memset(&ret, 0, sizeof(ret)); + + if (call(domain->conn, priv, 0, REMOTE_PROC_DOMAIN_GET_BLOCK_IO_THROTTLE, + (xdrproc_t) xdr_remote_domain_get_block_io_throttle_args, + (char *) &args, + (xdrproc_t) xdr_remote_domain_get_block_io_throttle_ret, + (char *) &ret) == -1) { + goto done; + } + + if (ret.found) { + reply->total_bytes_sec = ret.bps; + reply->read_bytes_sec = ret.bps_rd; + reply->write_bytes_sec = ret.bps_wr; + reply->total_iops_sec = ret.iops; + reply->read_iops_sec = ret.iops_rd; + reply->write_iops_sec = ret.iops_wr; + } + rv = 0; + +done: + remoteDriverUnlock(priv); + return rv; +} + /*----------------------------------------------------------------------*/ static virDrvOpenStatus ATTRIBUTE_NONNULL (1) @@ -4526,6 +4604,8 @@ static virDriver remote_driver = { .domainGetBlockJobInfo = remoteDomainGetBlockJobInfo, /* 0.9.4 */ .domainBlockJobSetSpeed = remoteDomainBlockJobSetSpeed, /* 0.9.4 */ .domainBlockPull = remoteDomainBlockPull, /* 0.9.4 */ + .domainSetBlockIoTune = remoteDomainSetBlockIoTune, /* 0.9.8 */ + .domainGetBlockIoTune = remoteDomainGetBlockIoTune, /* 0.9.8 */ }; static virNetworkDriver network_driver = { diff --git a/src/remote/remote_protocol.x b/src/remote/remote_protocol.x index a174af8..ccc77b5 100644 --- a/src/remote/remote_protocol.x +++ b/src/remote/remote_protocol.x @@ -1073,6 +1073,40 @@ struct remote_domain_block_pull_args { unsigned int flags; }; +struct remote_domain_set_block_io_throttle_args { + remote_nonnull_domain dom; + remote_nonnull_string disk; + unsigned hyper bps; + unsigned hyper bps_rd; + unsigned hyper bps_wr; + unsigned hyper iops; + unsigned hyper iops_rd; + unsigned hyper iops_wr; + unsigned int flags; +}; + +struct remote_domain_get_block_io_throttle_args { + remote_nonnull_domain dom; + remote_nonnull_string disk; + unsigned hyper bps; + unsigned hyper bps_rd; + unsigned hyper bps_wr; + unsigned hyper iops; + unsigned hyper iops_rd; + unsigned hyper iops_wr; + unsigned int flags; +}; + +struct remote_domain_get_block_io_throttle_ret { + unsigned hyper bps; + unsigned hyper bps_rd; + unsigned hyper bps_wr; + unsigned hyper iops; + unsigned hyper iops_rd; + unsigned hyper iops_wr; + unsigned int found; +}; + /* Network calls: */ struct remote_num_of_networks_ret { @@ -2562,7 +2596,9 @@ enum remote_procedure { REMOTE_PROC_DOMAIN_SNAPSHOT_NUM_CHILDREN = 246, /* autogen autogen priority:high */ REMOTE_PROC_DOMAIN_SNAPSHOT_LIST_CHILDREN_NAMES = 247, /* autogen autogen priority:high */ REMOTE_PROC_DOMAIN_EVENT_DISK_CHANGE = 248, /* skipgen skipgen */ - REMOTE_PROC_DOMAIN_OPEN_GRAPHICS = 249 /* skipgen skipgen */ + REMOTE_PROC_DOMAIN_OPEN_GRAPHICS = 249, /* skipgen skipgen */ + REMOTE_PROC_DOMAIN_SET_BLOCK_IO_THROTTLE = 250, /* skipgen skipgen */ + REMOTE_PROC_DOMAIN_GET_BLOCK_IO_THROTTLE = 251 /* skipgen skipgen */ /* * Notice how the entries are grouped in sets of 10 ? diff --git a/src/remote_protocol-structs b/src/remote_protocol-structs index 12cedef..4a813e0 100644 --- a/src/remote_protocol-structs +++ b/src/remote_protocol-structs @@ -757,6 +757,37 @@ struct remote_domain_block_pull_args { uint64_t bandwidth; u_int flags; }; +struct remote_domain_set_block_io_throttle_args { + remote_nonnull_domain dom; + remote_nonnull_string disk; + uint64_t bps; + uint64_t bps_rd; + uint64_t bps_wr; + uint64_t iops; + uint64_t iops_rd; + uint64_t iops_wr; + u_int flags; +}; +struct remote_domain_get_block_io_throttle_args { + remote_nonnull_domain dom; + remote_nonnull_string disk; + uint64_t bps; + uint64_t bps_rd; + uint64_t bps_wr; + uint64_t iops; + uint64_t iops_rd; + uint64_t iops_wr; + u_int flags; +}; +struct remote_domain_get_block_io_throttle_ret { + uint64_t bps; + uint64_t bps_rd; + uint64_t bps_wr; + uint64_t iops; + uint64_t iops_rd; + uint64_t iops_wr; + u_int found; +}; struct remote_num_of_networks_ret { int num; }; @@ -2005,4 +2036,6 @@ enum remote_procedure { REMOTE_PROC_DOMAIN_SNAPSHOT_LIST_CHILDREN_NAMES = 247, REMOTE_PROC_DOMAIN_EVENT_DISK_CHANGE = 248, REMOTE_PROC_DOMAIN_OPEN_GRAPHICS = 249, + REMOTE_PROC_DOMAIN_SET_BLOCK_IO_THROTTLE = 250, + REMOTE_PROC_DOMAIN_GET_BLOCK_IO_THROTTLE = 251, }; -- 1.7.1

2 1

[libvirt] [PATCH 1/8] Add new API virDomain{Set, Get}BlockIoTune
by Lei Li 10 Nov '11

10 Nov '11

This patch add new pulic API virDomainSetBlockIoTune and virDomainGetBlockIoTune. Signed-off-by: Zhi Yong Wu <wuzhy(a)linux.vnet.ibm.com> Signed-off-by: Lei Li <lilei(a)linux.vnet.ibm.com> --- include/libvirt/libvirt.h.in | 26 +++++++++ src/driver.h | 19 +++++++ src/libvirt.c | 115 ++++++++++++++++++++++++++++++++++++++++++ src/libvirt_public.syms | 2 + 4 files changed, 162 insertions(+), 0 deletions(-) diff --git a/include/libvirt/libvirt.h.in b/include/libvirt/libvirt.h.in index aa320b6..a79c35e 100644 --- a/include/libvirt/libvirt.h.in +++ b/include/libvirt/libvirt.h.in @@ -1640,6 +1640,32 @@ int virDomainBlockJobSetSpeed(virDomainPtr dom, const char *path, int virDomainBlockPull(virDomainPtr dom, const char *path, unsigned long bandwidth, unsigned int flags); +/* + * Block I/O throttling support + */ + +typedef struct _virDomainBlockIoTuneInfo virDomainBlockIoTuneInfo; +struct _virDomainBlockIoTuneInfo { + unsigned long long total_bytes_sec; + unsigned long long read_bytes_sec; + unsigned long long write_bytes_sec; + unsigned long long total_iops_sec; + unsigned long long read_iops_sec; + unsigned long long write_iops_sec; +}; +typedef virDomainBlockIoTuneInfo *virDomainBlockIoTuneInfoPtr; + +int +virDomainSetBlockIoTune(virDomainPtr dom, + const char *disk, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags); +int +virDomainGetBlockIoTune(virDomainPtr dom, + const char *disk, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags); + /* * NUMA support diff --git a/src/driver.h b/src/driver.h index 4c14aaa..9628ad7 100644 --- a/src/driver.h +++ b/src/driver.h @@ -741,6 +741,23 @@ typedef int unsigned long bandwidth, unsigned int flags); +/* + * Block I/O throttling support + */ + +typedef int + (*virDrvDomainSetBlockIoTune)(virDomainPtr dom, + const char *disk, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags); + +typedef int + (*virDrvDomainGetBlockIoTune)(virDomainPtr dom, + const char *disk, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags); + + /** * _virDriver: * @@ -899,6 +916,8 @@ struct _virDriver { virDrvDomainGetBlockJobInfo domainGetBlockJobInfo; virDrvDomainBlockJobSetSpeed domainBlockJobSetSpeed; virDrvDomainBlockPull domainBlockPull; + virDrvDomainSetBlockIoTune domainSetBlockIoTune; + virDrvDomainGetBlockIoTune domainGetBlockIoTune; }; typedef int diff --git a/src/libvirt.c b/src/libvirt.c index b0d1e01..79ac84d 100644 --- a/src/libvirt.c +++ b/src/libvirt.c @@ -17083,3 +17083,118 @@ error: virDispatchError(dom->conn); return -1; } + +/** + * virDomainSetBlockIoTune: + * @dom: pointer to domain object + * @disk: Fully-qualified disk name + * @info: Specify block I/O limits in bytes + * @flags: An OR'ed set of virDomainModificationImpact + * + * This function is mainly to enable Block I/O throttling function in libvirt. + * It is used to change the block I/O throttling setting for specified domain. + * + * Returns 0 if the operation has started, -1 on failure. + */ +int virDomainSetBlockIoTune(virDomainPtr dom, + const char *disk, + virDomainBlockIoTuneInfoPtr info, + unsigned int flags) +{ + virConnectPtr conn; + + VIR_DOMAIN_DEBUG(dom, "disk=%p, info=%p, flags=%x", + disk, info, flags); + + virResetLastError(); + + if (!VIR_IS_CONNECTED_DOMAIN (dom)) { + virLibDomainError(VIR_ERR_INVALID_DOMAIN, __FUNCTION__); + virDispatchError(NULL); + return -1; + } + conn = dom->conn; + + if (dom->conn->flags & VIR_CONNECT_RO) { + virLibDomainError(VIR_ERR_OPERATION_DENIED, __FUNCTION__); + goto error; + } + + if (!disk) { + virLibDomainError(VIR_ERR_INVALID_ARG, __FUNCTION__); + goto error; + } + + if (!info) { + virLibDomainError(VIR_ERR_INVALID_ARG, __FUNCTION__); + goto error; + } + + if (conn->driver->domainSetBlockIoTune) { + int ret; + ret = conn->driver->domainSetBlockIoTune(dom, disk, info, flags); + if (ret < 0) + goto error; + return ret; + } + + virLibDomainError(VIR_ERR_NO_SUPPORT, __FUNCTION__); + +error: + virDispatchError(dom->conn); + return -1; +} + +/** + * virDomainGetBlockIoTune: + * @dom: pointer to domain object + * @disk: Fully-qualified disk name + * @reply: Specify block I/O info in bytes + * @flags: An OR'ed set of virDomainModificationImpact + * + * This function is mainly to enable Block I/O throttling function in libvirt. + * It is used to get the block I/O throttling setting for specified domain. + * + * Returns 0 if the operation has started, -1 on failure. + */ + +int virDomainGetBlockIoTune(virDomainPtr dom, + const char *disk, + virDomainBlockIoTuneInfoPtr reply, + unsigned int flags) +{ + virConnectPtr conn; + + VIR_DOMAIN_DEBUG(dom, "disk=%p, reply=%p, flags=%x", + disk, reply, flags); + + virResetLastError(); + + if (!VIR_IS_CONNECTED_DOMAIN (dom)) { + virLibDomainError(VIR_ERR_INVALID_DOMAIN, __FUNCTION__); + virDispatchError(NULL); + return -1; + } + conn = dom->conn; + + if (!disk) { + virLibDomainError(VIR_ERR_INVALID_ARG, __FUNCTION__); + goto error; + } + + if (conn->driver->domainGetBlockIoTune) { + int ret; + ret = conn->driver->domainGetBlockIoTune(dom, disk, reply, flags); + if (ret < 0) + goto error; + return ret; + } + + virLibDomainError(VIR_ERR_NO_SUPPORT, __FUNCTION__); + +error: + virDispatchError(dom->conn); + return -1; + +} + diff --git a/src/libvirt_public.syms b/src/libvirt_public.syms index bcefb10..4808891 100644 --- a/src/libvirt_public.syms +++ b/src/libvirt_public.syms @@ -496,6 +496,8 @@ LIBVIRT_0.9.7 { virDomainSnapshotGetParent; virDomainSnapshotListChildrenNames; virDomainSnapshotNumChildren; + virDomainSetBlockIoTune; + virDomainGetBlockIoTune; } LIBVIRT_0.9.5; # .... define new API here using predicted next version number .... -- 1.7.1

4 5

[libvirt] qemu: Remove code instantiating filters on direct interfaces
by Stefan Berger 10 Nov '11

10 Nov '11

Remove code that instantiates network filters on direct type of interfaces. The parser already does not accept filters on those type of interfaces. --- src/qemu/qemu_command.c | 26 +------------------------- src/qemu/qemu_command.h | 1 - src/qemu/qemu_hotplug.c | 2 +- 3 files changed, 2 insertions(+), 27 deletions(-) Index: libvirt-acl/src/qemu/qemu_command.c =================================================================== --- libvirt-acl.orig/src/qemu/qemu_command.c +++ libvirt-acl/src/qemu/qemu_command.c @@ -135,7 +135,6 @@ uname_normalize (struct utsname *ut) */ int qemuPhysIfaceConnect(virDomainDefPtr def, - virConnectPtr conn, struct qemud_driver *driver, virDomainNetDefPtr net, virBitmapPtr qemuCaps, @@ -145,7 +144,6 @@ qemuPhysIfaceConnect(virDomainDefPtr def #if WITH_MACVTAP char *res_ifname = NULL; int vnet_hdr = 0; - int err; if (qemuCapsGet(qemuCaps, QEMU_CAPS_VNET_HDR) && net->model && STREQ(net->model, "virtio")) @@ -165,28 +163,6 @@ qemuPhysIfaceConnect(virDomainDefPtr def net->ifname = res_ifname; } - if (rc >=0 && driver->macFilter) { - if ((err = networkAllowMacOnPort(driver, net->ifname, net->mac))) { - virReportSystemError(err, - _("failed to add ebtables rule to allow MAC address on '%s'"), - net->ifname); - } - } - - if (rc >= 0) { - if ((net->filter) && (net->ifname)) { - err = virDomainConfNWFilterInstantiate(conn, net); - if (err) { - VIR_FORCE_CLOSE(rc); - delMacvtap(net->ifname, net->mac, - virDomainNetGetActualDirectDev(net), - virDomainNetGetActualDirectMode(net), - virDomainNetGetActualDirectVirtPortProfile(net), - driver->stateDir); - VIR_FREE(net->ifname); - } - } - } #else (void)def; (void)conn; @@ -4173,7 +4149,7 @@ qemuBuildCommandLine(virConnectPtr conn, tapfd) >= sizeof(tapfd_name)) goto no_memory; } else if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) { - int tapfd = qemuPhysIfaceConnect(def, conn, driver, net, + int tapfd = qemuPhysIfaceConnect(def, driver, net, qemuCaps, vmop); if (tapfd < 0) goto error; Index: libvirt-acl/src/qemu/qemu_command.h =================================================================== --- libvirt-acl.orig/src/qemu/qemu_command.h +++ libvirt-acl/src/qemu/qemu_command.h @@ -132,7 +132,6 @@ int qemuNetworkIfaceConnect(virDomainDef ATTRIBUTE_NONNULL(2); int qemuPhysIfaceConnect(virDomainDefPtr def, - virConnectPtr conn, struct qemud_driver *driver, virDomainNetDefPtr net, virBitmapPtr qemuCaps, Index: libvirt-acl/src/qemu/qemu_hotplug.c =================================================================== --- libvirt-acl.orig/src/qemu/qemu_hotplug.c +++ libvirt-acl/src/qemu/qemu_hotplug.c @@ -676,7 +676,7 @@ int qemuDomainAttachNetDevice(virConnect if (qemuOpenVhostNet(vm->def, net, priv->qemuCaps, &vhostfd) < 0) goto cleanup; } else if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) { - if ((tapfd = qemuPhysIfaceConnect(vm->def, conn, driver, net, + if ((tapfd = qemuPhysIfaceConnect(vm->def, driver, net, priv->qemuCaps, VIR_VM_OP_CREATE)) < 0) goto cleanup;

2 2

[libvirt] [PATCH 0/4] Support macvlan devices for LXC containers
by Daniel P. Berrange 10 Nov '11

10 Nov '11

This series does the bare minimum required to support the network inteface type=direct for LXC containers, using macvlan devices as the implementation. There is a slight complication though... For bridged container NICs we create a veth pair of devices. One of the devices lives host side, and is enslaved in a bridge. The other veth device gets moved to the container namespace to form the eth0. So we both both a host & container side device visible. For direct container NICS we create a macvlan device, which is moved to the container namespace. There is no host side interface that is dedicated for the container - only the general ethernet device the macvlan is bound to. Since there is no host side interface for the container we are unable to create network filter rules, or network bandwidth controls. In addition while we could perform the 8021.Qb{gh} association during container startup, before moving the macvlan device to the container namespace, we can't perform any disassociation on container shutdown. By the time we see the container has shutdown, the macvlan device has already been killed off. The inability to setup iptables/tc rules against devices that are only visible in the container namespace is arguably a flaw in the Linux kernel's namespace code support for iptables/tc. There ought to be a syntax for iptables/tc to write rules which affect NICs in other namespaces

2 10

[libvirt] [PATCH 0/5] Support Blkio tune, CPU tune, NUMA, CPU affinity in LXC
by Daniel P. Berrange 10 Nov '11

10 Nov '11

The following patch series brings the LXC driver upto parity with the QEMU driver in terms of support for various performance tuning controls

1 5

[libvirt] [PATCH] Disable numactl on ARM architectures too
by Daniel P. Berrange 10 Nov '11

10 Nov '11

From: "Daniel P. Berrange" <berrange(a)redhat.com> Pushed under trivial rule - a backport from the Fedora spec * libvirt.spec.in: Disable numactl on ARM --- libvirt.spec.in | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-) diff --git a/libvirt.spec.in b/libvirt.spec.in index 10280f0..d4e3e17 100644 --- a/libvirt.spec.in +++ b/libvirt.spec.in @@ -92,8 +92,8 @@ %define with_libxl 0 %endif -# Numactl is not available on s390[x] -%ifarch s390 s390x +# Numactl is not available on s390[x] and ARM +%ifarch s390 s390x %{arm} %define with_numactl 0 %endif -- 1.7.6.4

1 0

[libvirt] [PATCH] Add libvirt confdir to files section in mingw32 spec
by Daniel P. Berrange 10 Nov '11

10 Nov '11

From: "Daniel P. Berrange" <berrange(a)redhat.com> Pushed under the trivial rule - this fix pulled back from Fedora * mingw32-libvirt.spec.in: Ensure we own the confdir --- mingw32-libvirt.spec.in | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/mingw32-libvirt.spec.in b/mingw32-libvirt.spec.in index c2690f3..89d1d7f 100644 --- a/mingw32-libvirt.spec.in +++ b/mingw32-libvirt.spec.in @@ -126,6 +126,7 @@ rm -rf $RPM_BUILD_ROOT %files %defattr(-,root,root) +%dir %{_mingw32_sysconfdir}/libvirt/ %config(noreplace) %{_mingw32_sysconfdir}/libvirt/libvirt.conf %{_mingw32_bindir}/libvirt-0.dll -- 1.7.6.4

1 0

[libvirt] nwfilter - limit VM traffic to specific mac address
by Shahar Havivi 10 Nov '11

10 Nov '11

Hi, I want to limit VM traffic to a specific MAC address, ie VMs cannot traffic each other other then a specific gateway. I am using custom nwfilter name: isolatedprivatevlan-vdsm.xml located in /etc/libvirt/nwfilter/: <filter name='isolatedprivatevlan-vdsm' chain='root'> <filterref filter='clean-traffic'/> <rule action='drop' direction='out' priority='500'> <mac match='no' dstmacaddr='$GATEWAY_MAC'/> </rule> </filter> VM1 domian xml portion: <interface type="bridge"> <mac address="00:1a:4a:16:01:53"/> <model type="virtio"/> <source bridge="red"/> <filterref filter="isolatedprivatevlan-vdsm"> <parameter name="GATEWAY_MAC" value="00:00:0c:07:ac:00"/> </filterref> </interface> VM2 domian xml portion: <interface type="bridge"> <mac address="00:1a:4a:16:01:52"/> <model type="virtio"/> <source bridge="red"/> <filterref filter="isolatedprivatevlan-vdsm"> <parameter name="GATEWAY_MAC" value="00:00:0c:07:ac:00"/> </filterref> </interface> in each VM (Fedora 15 LiveCD) I assign ip: # ifconfig eth0 10.35.1.240 netmask 255.255.254.0 # route add default gw 10.35.1.1 vm2: # ifconfig eth0 10.35.1.241 netmask 255.255.254.0 # route add default gw 10.35.1.1 but the filter is not working, I can ping the VMs from each other, Am I missing something? Thanks, Shahar Havivi.

2 11

Re: [libvirt] [PATCH] fix crash when starting network
by lvroyce 10 Nov '11

10 Nov '11

tested-by: Wen Ruo Lv<lvroyce(a)linux.vnet.ibm.com> tested 1.net-start cmd with bridge mac specified (failed) 2.attach-device cmd with a hot plug nic(ok) My network is as below: <network> <name>default</name> <uuid>361441af-e1f0-472d-a503-dfcbbefa03fb</uuid> <forward mode='nat'/> <bridge name='virbr0' stp='on' delay='0' /> <mac address='00:16:3E:5D:C7:9E'/> <ip address='192.168.122.1' netmask='255.255.255.0'> <dhcp> <range start='192.168.122.2' end='192.168.122.254' /> </dhcp> </ip> </network> It fails at brSetInterfaceMac-->return ioctl(ctl->fd, SIOCSIFHWADDR, &ifr) but SIOCGIFHWADDR succeed. error msg is: cannot create dummy tap device 'virbr0-nic' to set mac address on bridge 'virbr0':no such device

1 0

[libvirt] [test-API][PATCH 1/3] add new "clean" keyword to clean environment after each testcase
by Guannan Ren 10 Nov '11

10 Nov '11

testcase1 option1 avalue option2 bvalue clean testcase2 option1 cvalue option2 dvalue clean --- generator.py | 44 ++++++++++++++++++++++++++++++++------------ 1 files changed, 32 insertions(+), 12 deletions(-) diff --git a/generator.py b/generator.py index 5a3a2ab..2d59353 100644 --- a/generator.py +++ b/generator.py @@ -41,6 +41,7 @@ class FuncGen(object): self.lockfile = lockfile self.bugstxt = bugstxt self.loglevel = loglevel + self.testcase_number = 0 self.fmt = format.Format(logfile) self.log_xml_parser = log_xml_parser @@ -49,19 +50,21 @@ class FuncGen(object): self.__case_info_save(activity, testrunid) mapper_obj = mapper.Mapper(activity) - pkg_tripped_cases = mapper_obj.get_package_tripped() + pkg_casename_func = mapper_obj.package_casename_func_map() - for test_procedure in pkg_tripped_cases: + for test_procedure in pkg_casename_func: log_xml_parser.add_testprocedure_xml(testrunid, testid, test_procedure) self.cases_ref_names = [] - for case in pkg_tripped_cases: + for case in pkg_casename_func: case_ref_name = case.keys()[0] + if case_ref_name[-6:] != "_clean": + self.testcase_number += 1 self.cases_ref_names.append(case_ref_name) self.cases_params_list = [] - for case in pkg_tripped_cases: + for case in pkg_casename_func: case_params = case.values()[0] self.cases_params_list.append(case_params) @@ -101,7 +104,7 @@ class FuncGen(object): envlog = log.EnvLog(self.logfile, self.loglevel) logger = envlog.env_log() - testcase_number = len(self.cases_ref_names) + loop_number = len(self.cases_ref_names) start_time = time.strftime("%Y-%m-%d %H:%M:%S") logger.info("Checking Testing Environment... ") @@ -111,7 +114,7 @@ class FuncGen(object): sys.exit(1) else: logger.info("\nStart Testing:") - logger.info(" Case Count: %s" % testcase_number) + logger.info(" Case Count: %s" % self.testcase_number) logger.info(" Log File: %s\n" % self.logfile) del envlog @@ -119,21 +122,31 @@ class FuncGen(object): logger = caselog.case_log() retflag = 0 - for i in range(testcase_number): + for i in range(loop_number): case_ref_name = self.cases_ref_names[i] - self.fmt.printf('start', case_ref_name) + pkg_casename = case_ref_name.rpartition(":")[0] + funcname = case_ref_name.rpartition(":")[-1] + + cleanoper = 0 if "_clean" not in funcname else 1 + + if not cleanoper: + self.fmt.printf('start', pkg_casename) + else: + self.fmt.printf('string', 12*" " + "Cleaning...") + case_params = self.cases_params_list[i] case_start_time = time.strftime("%Y-%m-%d %H:%M:%S") ret = -1 + clean_ret = -1 try: try: if case_ref_name != 'sleep': case_params['logger'] = logger - existed_bug_list = self.bug_check(case_ref_name) + existed_bug_list = self.bug_check(pkg_casename) if len(existed_bug_list) == 0: if case_ref_name == 'sleep': @@ -143,13 +156,16 @@ class FuncGen(object): ret = 0 else: ret = self.cases_func_ref_dict[case_ref_name](case_params) + if cleanoper: + clean_ret = ret + ret = 0 else: logger.info("about the testcase , bug existed:") for existed_bug in existed_bug_list: logger.info("%s" % existed_bug) ret = 100 - self.fmt.printf('end', case_ref_name, ret) + self.fmt.printf('end', pkg_casename, ret) continue except Exception, e: logger.error(traceback.format_exc()) @@ -163,7 +179,11 @@ class FuncGen(object): else: pass retflag += ret - self.fmt.printf('end', case_ref_name, ret) + + if not cleanoper: + self.fmt.printf('end', pkg_casename, ret) + else: + self.fmt.printf('string', 21*" " + "Done" if clean_ret < 1 else 21*" " + "Fail") end_time = time.strftime("%Y-%m-%d %H:%M:%S") del caselog @@ -172,7 +192,7 @@ class FuncGen(object): logger = envlog.env_log() logger.info("\nSummary:") logger.info(" Total:%s [Pass:%s Fail:%s]" % \ - (testcase_number, (testcase_number - retflag), retflag)) + (self.testcase_number, (self.testcase_number - retflag), retflag)) del envlog result = (retflag and "FAIL") or "PASS" -- 1.7.1

2 4

[libvirt] [PATCH 0/2] fix nwfilter when /tmp is mounted noexec
by Eric Blake 10 Nov '11

10 Nov '11

https://bugzilla.redhat.com/show_bug.cgi?id=752254 points out that libvirt cannot support nwfilter on a system with /tmp mounted noexec (which is a very common setup in security-conscious setups), all because we were trying to directly invoke a temporary script instead of invoking a shell to read the script. I've split this patch into 2 parts, on the off-chance that patch 2 would run afoul of command line length limits (if the total size of the generated nwfilter commands could possibly cause E2BIG, then we have to go through a temporary file). But my recollection is that modern Linux kernels support unlimited command-line length (that is, ARG_MAX is not a concern on Linux), and that nwfilter_ebiptables_driver only compiles on Linux, so my preference would be to squash these into a single commit, if others agree that we don't have to worry about length limits. At any rate, I'm quite impressed at the number of lines of code I was able to remove in order to fix a bug! Eric Blake (2): nwfilter: avoid failure with noexec /tmp nwfilter: simplify execution of ebiptables scripts src/nwfilter/nwfilter_ebiptables_driver.c | 134 ++-------------------------- 1 files changed, 10 insertions(+), 124 deletions(-) -- 1.7.4.4

3 10

[libvirt] [PATCH] command: handle empty buffer argument correctly
by Eric Blake 10 Nov '11

10 Nov '11

virBufferContentAndReset (intentionally) returns NULL for a buffer with no content, but it is feasible to invoke a command with an explicit empty string. * src/util/command.c (virCommandAddEnvBuffer): Reject empty string. (virCommandAddArgBuffer): Allow explicit empty argument. * tests/commandtest.c (test9): Test it. * tests/commanddata/test9.log: Adjust. --- As pointed out here: https://www.redhat.com/archives/libvir-list/2011-November/msg00435.html src/util/command.c | 13 ++++++++++++- tests/commanddata/test9.log | 4 +++- tests/commandtest.c | 13 ++++++++++++- 3 files changed, 27 insertions(+), 3 deletions(-) diff --git a/src/util/command.c b/src/util/command.c index c3ce361..f5effdf 100644 --- a/src/util/command.c +++ b/src/util/command.c @@ -983,6 +983,10 @@ virCommandAddEnvBuffer(virCommandPtr cmd, virBufferPtr buf) virBufferFreeAndReset(buf); return; } + if (!virBufferUse(buf)) { + cmd->has_error = EINVAL; + return; + } cmd->env[cmd->nenv++] = virBufferContentAndReset(buf); } @@ -1092,7 +1096,14 @@ virCommandAddArgBuffer(virCommandPtr cmd, virBufferPtr buf) return; } - cmd->args[cmd->nargs++] = virBufferContentAndReset(buf); + cmd->args[cmd->nargs] = virBufferContentAndReset(buf); + if (!cmd->args[cmd->nargs]) + cmd->args[cmd->nargs] = strdup(""); + if (!cmd->args[cmd->nargs]) { + cmd->has_error = ENOMEM; + return; + } + cmd->nargs++; } diff --git a/tests/commanddata/test9.log b/tests/commanddata/test9.log index 2607530..3a93c19 100644 --- a/tests/commanddata/test9.log +++ b/tests/commanddata/test9.log @@ -2,8 +2,10 @@ ARG:-version ARG:-log=bar.log ARG:arg1 ARG:arg2 -ARG:arg3 +ARG: ARG:arg4 +ARG:arg5 +ARG:arg6 ENV:DISPLAY=:0.0 ENV:HOME=/home/test ENV:HOSTNAME=test diff --git a/tests/commandtest.c b/tests/commandtest.c index dd6c248..efc48fe 100644 --- a/tests/commandtest.c +++ b/tests/commandtest.c @@ -352,11 +352,22 @@ static int test9(const void *unused ATTRIBUTE_UNUSED) { virCommandPtr cmd = virCommandNew(abs_builddir "/commandhelper"); const char* const args[] = { "arg1", "arg2", NULL }; + virBuffer buf = VIR_BUFFER_INITIALIZER; virCommandAddArg(cmd, "-version"); virCommandAddArgPair(cmd, "-log", "bar.log"); virCommandAddArgSet(cmd, args); - virCommandAddArgList(cmd, "arg3", "arg4", NULL); + virCommandAddArgBuffer(cmd, &buf); + virBufferAddLit(&buf, "arg4"); + virCommandAddArgBuffer(cmd, &buf); + virCommandAddArgList(cmd, "arg5", "arg6", NULL); + + if (virBufferUse(&buf)) { + printf("Buffer not transferred\n"); + virBufferFreeAndReset(&buf); + virCommandFree(cmd); + return -1; + } if (virCommandRun(cmd, NULL) < 0) { virErrorPtr err = virGetLastError(); -- 1.7.4.4

1 0

[libvirt] [libvirt-glib 1/3] Add gvir_config_object_new{_from_xml} helpers
by Christophe Fergeau 09 Nov '11

09 Nov '11

This factors the code that is repeated in the _new methods of all classes deriving from GVirObject --- libvirt-gconfig/libvirt-gconfig-domain.c | 26 ++++++++++-------------- libvirt-gconfig/libvirt-gconfig-object.c | 31 ++++++++++++++++++++++++++++++ libvirt-gconfig/libvirt-gconfig-object.h | 8 +++++++ libvirt-gconfig/libvirt-gconfig.sym | 1 + 4 files changed, 51 insertions(+), 15 deletions(-) diff --git a/libvirt-gconfig/libvirt-gconfig-domain.c b/libvirt-gconfig/libvirt-gconfig-domain.c index 837287c..3290389 100644 --- a/libvirt-gconfig/libvirt-gconfig-domain.c +++ b/libvirt-gconfig/libvirt-gconfig-domain.c @@ -147,27 +147,23 @@ static void gvir_config_domain_init(GVirConfigDomain *conn) GVirConfigDomain *gvir_config_domain_new_from_xml(const gchar *xml, GError **error) { - xmlNodePtr node; + GVirConfigObject *object; - node = gvir_config_xml_parse(xml, "domain", error); - if ((error != NULL) && (*error != NULL)) - return NULL; - return GVIR_CONFIG_DOMAIN(g_object_new(GVIR_TYPE_CONFIG_DOMAIN, - "node", node, - "schema", DATADIR "/libvirt/schemas/domain.rng", - NULL)); + object = gvir_config_object_new_from_xml(GVIR_TYPE_CONFIG_DOMAIN, + "domain", + DATADIR "/libvirt/schemas/domain.rng", + xml, error); + return GVIR_CONFIG_DOMAIN(object); } GVirConfigDomain *gvir_config_domain_new(void) { - xmlDocPtr doc; + GVirConfigObject *object; - doc = xmlNewDoc((xmlChar *)"1.0"); - doc->children = xmlNewDocNode(doc, NULL, (xmlChar *)"domain", NULL); - return GVIR_CONFIG_DOMAIN(g_object_new(GVIR_TYPE_CONFIG_DOMAIN, - "node", doc->children, - "schema", DATADIR "/libvirt/schemas/domain.rng", - NULL)); + object = gvir_config_object_new(GVIR_TYPE_CONFIG_DOMAIN, + "domain", + DATADIR "/libvirt/schemas/domain.rng"); + return GVIR_CONFIG_DOMAIN(object); } char *gvir_config_domain_get_name(GVirConfigDomain *domain) diff --git a/libvirt-gconfig/libvirt-gconfig-object.c b/libvirt-gconfig/libvirt-gconfig-object.c index e59cbcd..adbd955 100644 --- a/libvirt-gconfig/libvirt-gconfig-object.c +++ b/libvirt-gconfig/libvirt-gconfig-object.c @@ -355,3 +355,34 @@ void gvir_config_object_set_node_content_uint64(GVirConfigObject *object, gvir_config_object_set_node_content(object, node_name, str); g_free(str); } + +GVirConfigObject *gvir_config_object_new_from_xml(GType type, + const char *root_name, + const char *schema, + const gchar *xml, + GError **error) +{ + xmlNodePtr node; + + node = gvir_config_xml_parse(xml, root_name, error); + if ((error != NULL) && (*error != NULL)) + return NULL; + return GVIR_CONFIG_OBJECT(g_object_new(type, + "node", node, + "schema", schema, + NULL)); +} + +GVirConfigObject *gvir_config_object_new(GType type, + const char *root_name, + const char *schema) +{ + xmlDocPtr doc; + + doc = xmlNewDoc((xmlChar *)"1.0"); + doc->children = xmlNewDocNode(doc, NULL, (xmlChar *)root_name, NULL); + return GVIR_CONFIG_OBJECT(g_object_new(type, + "node", doc->children, + "schema", schema, + NULL)); +} diff --git a/libvirt-gconfig/libvirt-gconfig-object.h b/libvirt-gconfig/libvirt-gconfig-object.h index c203798..52e4525 100644 --- a/libvirt-gconfig/libvirt-gconfig-object.h +++ b/libvirt-gconfig/libvirt-gconfig-object.h @@ -59,6 +59,14 @@ struct _GVirConfigObjectClass GType gvir_config_object_get_type(void); +GVirConfigObject *gvir_config_object_new(GType type, + const char *root_name, + const char *schema); +GVirConfigObject *gvir_config_object_new_from_xml(GType type, + const char *root_name, + const char *schema, + const gchar *xml, + GError **error); void gvir_config_object_validate(GVirConfigObject *config, GError **err); diff --git a/libvirt-gconfig/libvirt-gconfig.sym b/libvirt-gconfig/libvirt-gconfig.sym index 951aac6..31a89d2 100644 --- a/libvirt-gconfig/libvirt-gconfig.sym +++ b/libvirt-gconfig/libvirt-gconfig.sym @@ -31,6 +31,7 @@ LIBVIRT_GOBJECT_0.0.1 { gvir_config_object_get_type; gvir_config_object_error_quark; gvir_config_object_new; + gvir_config_object_new_from_xml; gvir_config_object_get_schema; gvir_config_object_get_xml_node; gvir_config_object_to_xml; -- 1.7.7

2 5

[libvirt] [PATCH 8/8] Add tests for blkdeviotune
by Lei Li 09 Nov '11

09 Nov '11

Signed-off-by: Zhi Yong Wu <wuzhy(a)linux.vnet.ibm.com> Signed-off-by: Lei Li <lilei(a)linux.vnet.ibm.com> --- .../qemuxml2argv-blkdeviotune.args | 4 ++ .../qemuxml2argvdata/qemuxml2argv-blkdeviotune.xml | 37 ++++++++++++++++++++ tests/qemuxml2argvtest.c | 1 + tests/qemuxml2xmltest.c | 1 + 4 files changed, 43 insertions(+), 0 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.xml diff --git a/tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.args b/tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.args new file mode 100644 index 0000000..6f34698 --- /dev/null +++ b/tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.args @@ -0,0 +1,4 @@ +LC_ALL=C PATH=/bin HOME=/home/test USER=test LOGNAME=test /usr/bin/qemu -S -M \ +pc -m 214 -smp 1 -name QEMUGuest1 -nographic -monitor unix:/tmp/test-monitor, \ +server,nowait -no-acpi -boot c -hda /dev/HostVG/QEMUGuest1 -net none -serial \ +none -parallel none -usb diff --git a/tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.xml b/tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.xml new file mode 100644 index 0000000..0cabb90 --- /dev/null +++ b/tests/qemuxml2argvdata/qemuxml2argv-blkdeviotune.xml @@ -0,0 +1,37 @@ +<domain type='qemu'> + <name>QEMUGuest1</name> + <uuid>c7a5fdbd-edaf-9455-926a-d65c16db1809</uuid> + <memory>219100</memory> + <currentMemory>219100</currentMemory> + <vcpu>1</vcpu> + <os> + <type arch='i686' machine='pc'>hvm</type> + <boot dev='hd'/> + </os> + <clock offset='utc'/> + <on_poweroff>destroy</on_poweroff> + <on_reboot>restart</on_reboot> + <on_crash>destroy</on_crash> + <devices> + <emulator>/usr/bin/qemu</emulator> + <disk type='block' device='disk' snapshot='internal'> + <driver name='qemu' type='qcow2' cache='none'/> + <source dev='/dev/HostVG/QEMUGuest1'/> + <target dev='hda' bus='ide'/> + <iotune> + <total_bytes_sec>5000</total_bytes_sec> + <total_iops_sec>6000</total_iops_sec> + <address type='drive' controller='0' bus='0' unit='0'/> + </disk> + <disk type='block' device='cdrom' snapshot='no'> + <driver name='qemu' type='raw'/> + <source dev='/dev/HostVG/QEMUGuest2'/> + <target dev='hdc' bus='ide'/> + <readonly/> + <address type='drive' controller='0' bus='1' unit='0'/> + </disk> + <controller type='ide' index='0'/> + <memballoon model='virtio'/> + </devices> +</domain> + diff --git a/tests/qemuxml2argvtest.c b/tests/qemuxml2argvtest.c index d9a6e8d..1ebb950 100644 --- a/tests/qemuxml2argvtest.c +++ b/tests/qemuxml2argvtest.c @@ -585,6 +585,7 @@ mymain(void) DO_TEST("blkiotune", false, QEMU_CAPS_NAME); DO_TEST("cputune", false, QEMU_CAPS_NAME); DO_TEST("numatune-memory", false, NONE); + DO_TEST("blkdeviotune", false, QEMU_CAPS_NAME); DO_TEST("multifunction-pci-device", false, QEMU_CAPS_DRIVE, QEMU_CAPS_DEVICE, QEMU_CAPS_NODEFCONFIG, diff --git a/tests/qemuxml2xmltest.c b/tests/qemuxml2xmltest.c index 3f37520..2e6b5c7 100644 --- a/tests/qemuxml2xmltest.c +++ b/tests/qemuxml2xmltest.c @@ -191,6 +191,7 @@ mymain(void) DO_TEST("event_idx"); DO_TEST("usb-redir"); + DO_TEST("blkdeviotune"); /* These tests generate different XML */ DO_TEST_DIFFERENT("balloon-device-auto"); -- 1.7.1

1 0

[libvirt] [PATCH 6/8] Doc: Add description and validate the new XML
by Lei Li 09 Nov '11

09 Nov '11

Signed-off-by: Zhi Yong Wu <wuzhy(a)linux.vnet.ibm.com> Signed-off-by: Lei Li <lilei(a)linux.vnet.ibm.com> --- docs/formatdomain.html.in | 31 +++++++++++++++++++++++++++++++ docs/schemas/domaincommon.rng | 24 ++++++++++++++++++++++++ 2 files changed, 55 insertions(+), 0 deletions(-) diff --git a/docs/formatdomain.html.in b/docs/formatdomain.html.in index cbad196..733062d 100644 --- a/docs/formatdomain.html.in +++ b/docs/formatdomain.html.in @@ -893,6 +893,14 @@ <driver name="tap" type="aio" cache="default"/> <source file='/var/lib/xen/images/fv0'/ startupPolicy='optional'> <target dev='hda' bus='ide'/> + <iotune> + <total_bytes_sec>n</total_bytes_sec> + <read_bytes_sec>n</read_bytes_sec> + <write_bytes_sec>n</write_bytes_sec> + <total_iops_sec>n</total_iops_sec> + <read_bytes_sec>n</read_bytes_sec> + <write_bytes_sec>n</write_bytes_sec> + </iotune> <boot order='2'/> <encryption type='...'> ... @@ -1010,6 +1018,29 @@ <span class="since">Since 0.0.3; <code>bus</code> attribute since 0.4.3; "usb" attribute value since after 0.4.4; "sata" attribute value since 0.9.7</span></dd> + <dt><code>iotune</code></dt> + <dd>The optional <code>iotune</code> element provides the ability + to set or get block I/O throttling for the device. Block I/O + throtting be implemented by qemu, is specified per-disk and can + vary across multiple disks.</dd> + <dt><code>total_bytes_sec</code></dt> + <dd>The optinal <code>total_bytes_sec</code> element is the total throughput + limit in bytes per second.</dd> + <dt><code>read_bytes_sec</code></dt> + <dd>The optinal <code>read_bytes_sec</code> element is the read throughput + limit in bytes per second.</dd> + <dt><code>write_bytes_sec</code</dt> + <dd>The optinal <code>write_bytes_sec</code> element is the write throughput + limit in bytes per second.</dd> + <dt><code>total_iops_sec</code></dt> + <dd>The optional <code>total_iops_sec</code> element is the total I/O operations + per second.</dd> + <dt><code>read_iops_sec</code></dt> + <dd>The optional <code>read_iops_sec</code> element is the read I/O operations + per second.</dd> + <dt><code>write_iops_sec</code></dt> + <dd>The optional <code>write_iops_sec</code> element is the write I/O operations + per second.</dd> <dt><code>driver</code></dt> <dd> The optional driver element allows specifying further details diff --git a/docs/schemas/domaincommon.rng b/docs/schemas/domaincommon.rng index b6f858e..c6873a0 100644 --- a/docs/schemas/domaincommon.rng +++ b/docs/schemas/domaincommon.rng @@ -698,6 +698,30 @@ <optional> <ref name="snapshot"/> </optional> + <optional> + <element name="iotune"> + <choice> + <element name="total_bytes_sec"> + <ref name="unsignedLongLong"> + </element> + <element name="read_bytes_sec"> + <ref name="unsignedLongLong"> + </element> + <element name="write_bytes_sec"> + <ref name="unsignedLongLong"> + </element> + <element name="total_iops_sec"> + <ref name="unsignedLongLong"> + </element> + <element name="read_iops_sec"> + <ref name="unsignedLongLong"> + </element> + <element name="write_iopw_sec"> + <ref name="unsignedLongLong"> + </emement> + </choice> + </element> + </optional> <choice> <group> <attribute name="type"> -- 1.7.1

1 0

[libvirt] Virtual serial logging server?
by Reeted 09 Nov '11

09 Nov '11

Dear all, please excuse the almost-OT question, I see various possibilities in quemu-kvm and libvirt for sending virtual serial port data to files, sockets, pipes, etc on the host. In particular, the TCP socket seems interesting. Can you suggest a server application to receive all such TCP connections and log serial data for many virtual machines at once? In particular I would be interested in something with quotas, i.e. something that deletes old lines from the logs of a certain VM when the filesystem space occupied by the serial logs of such VM gets over a certain amount of space. So that the log space for other VMs is not starved in case one of them loops. Thank you R.

4 3

[libvirt] [PATCH 0/2] Fix 0.9.7 for mingw
by Eric Blake 09 Nov '11

09 Nov '11

I'm proposing two options to fix the build failure mentioned here: https://www.redhat.com/archives/libvir-list/2011-November/msg00309.html without having to cut a release of 0.9.8. Note that this email is presenting a diff of a diff - that is, patch 2/2 is a diff that creates two new files, whose contents are then diffs themselves; it is those resulting diffs that get used in two contexts - the gnulib local diffs when building libvirt.git, and the application of those diffs to the non-version-controlled libvirt-0.9.7/gnulib/lib files when backporting the changes for building a mingw rpm. Hopefully that doesn't confuse you too much. Option 1: Upstream applies both of these patches, then the fix for building the mingw32-libvirt RPM would be to take the resulting diff files in patch 2/2 and apply those directly to the gnulib/lib files present in the 0.9.7 tarball expansion (no .gnulib submodule updated, and no gnulib-tool run introduced into the spec file). Later, when upstream gnulib is fixed, we revert patch 2/2 at the same time we update our .gnulib submodule. Option 2 would be to apply patch 1/2, but then wait until the fixes in patch 2/2 are present in upstream gnulib, and just update .gnulib directly instead of bothering with local diffs. Personally, I prefer option 1. I've tested this by running ./autobuild.sh, which builds both Linux and cross-builds to mingw, validating that the diffs apply correctly so that mingw builds. Eric Blake (2): build: allow for local gnulib diffs build: fix mingw build of gnulib openpty autogen.sh | 3 +++ bootstrap.conf | 6 ++++-- cfg.mk | 1 + gnulib/local/lib/openpty.c.diff | 26 ++++++++++++++++++++++++++ gnulib/local/lib/pty.in.h.diff | 13 +++++++++++++ 5 files changed, 47 insertions(+), 2 deletions(-) create mode 100644 gnulib/local/lib/openpty.c.diff create mode 100644 gnulib/local/lib/pty.in.h.diff -- 1.7.4.4

2 8

[libvirt] [PATCH] Allow non-blocking message sending on virNetClient
by Daniel P. Berrange 09 Nov '11

09 Nov '11

From: "Daniel P. Berrange" <berrange(a)redhat.com> Split the existing virNetClientSend into two parts virNetClientSend and virNetClientSendNoReply, instead of having a 'bool expectReply' parameter. Add a new virNetClientSendNonBlock which returns 2 on full send, 1 on partial send, 0 on no send, -1 on error If a partial send occurs, then a subsequent call to any of the virNetClientSend* APIs will finish any outstanding I/O. TODO: the virNetClientEvent event handler could be used to speed up completion of partial sends if an event loop is present. * src/rpc/virnetclientprogram.c, src/rpc/virnetclientstream.c: Update for changed API --- src/rpc/virnetclient.c | 249 +++++++++++++++++++++++++++++++++------- src/rpc/virnetclient.h | 12 ++- src/rpc/virnetclientprogram.c | 2 +- src/rpc/virnetclientstream.c | 11 ++- 4 files changed, 224 insertions(+), 50 deletions(-) diff --git a/src/rpc/virnetclient.c b/src/rpc/virnetclient.c index 4b7d4a9..b0ed507 100644 --- a/src/rpc/virnetclient.c +++ b/src/rpc/virnetclient.c @@ -55,6 +55,8 @@ struct _virNetClientCall { virNetMessagePtr msg; bool expectReply; + bool nonBlock; + bool haveThread; virCond cond; @@ -86,8 +88,15 @@ struct _virNetClient { int wakeupSendFD; int wakeupReadFD; - /* List of threads currently waiting for dispatch */ + /* + * List of calls currently waiting for dispatch + * The calls should all have threads waiting for + * them, except possibly the first call in the list + * which might be a partially sent non-blocking call. + */ virNetClientCallPtr waitDispatch; + /* Whether a thread is dispatching */ + bool haveTheBuck; size_t nstreams; virNetClientStreamPtr *streams; @@ -555,7 +564,7 @@ virNetClientCallDispatchReply(virNetClientPtr client) virNetClientCallPtr thecall; /* Ok, definitely got an RPC reply now find - out who's been waiting for it */ + out which waiting call is associated with it */ thecall = client->waitDispatch; while (thecall && !(thecall->msg->header.prog == client->msg.header.prog && @@ -896,10 +905,31 @@ virNetClientIOHandleInput(virNetClientPtr client) } +static void virNetClientPassTheBuck(virNetClientPtr client) +{ + virNetClientCallPtr tmp = client->waitDispatch; + + /* See if someone else is still waiting + * and if so, then pass the buck ! */ + while (tmp) { + if (tmp->haveThread) { + VIR_DEBUG("Passing the buck to %p", tmp); + virCondSignal(&tmp->cond); + return; + } + tmp = tmp->next; + } + VIR_DEBUG("No thread to pass the buck to"); +} + + /* * Process all calls pending dispatch/receive until we * get a reply to our own call. Then quit and pass the buck * to someone else. + * + * Returns 2 if fully sent, 1 if partially sent (only for nonBlock==true), + * 0 if nothing sent (only for nonBlock==true) and -1 on error */ static int virNetClientIOEventLoop(virNetClientPtr client, virNetClientCallPtr thiscall) @@ -924,6 +954,12 @@ static int virNetClientIOEventLoop(virNetClientPtr client, if (virNetSocketHasCachedData(client->sock)) timeout = 0; + /* If we're a non-blocking call, then we don't + * want to wait for I/O readyness + */ + if (thiscall->nonBlock) + timeout = 0; + fds[0].events = fds[0].revents = 0; fds[1].events = fds[1].revents = 0; @@ -975,8 +1011,34 @@ static int virNetClientIOEventLoop(virNetClientPtr client, /* If we have existing SASL decoded data, pretend * the socket became readable so we consume it */ - if (virNetSocketHasCachedData(client->sock)) + if (virNetSocketHasCachedData(client->sock)) { fds[0].revents |= POLLIN; + } else if (ret == 0 && thiscall->nonBlock) { + if (thiscall->msg->bufferOffset == 0) { + /* No data sent at all, remove ourselves from the list */ + tmp = client->waitDispatch; + prev = NULL; + while (tmp) { + if (tmp == thiscall) { + if (prev) { + prev->next = thiscall->next; + } else { + client->waitDispatch = thiscall->next; + } + break; + } + prev = tmp; + tmp = tmp->next; + } + VIR_DEBUG("Giving up the buck %p %p", thiscall, client->waitDispatch); + virNetClientPassTheBuck(client); + return 0; + } else { + VIR_DEBUG("Giving up the buck %p %p", thiscall, client->waitDispatch); + virNetClientPassTheBuck(client); + return 1; /* partial send */ + } + } if (fds[1].revents) { VIR_DEBUG("Woken up from poll by other thread"); @@ -988,6 +1050,7 @@ static int virNetClientIOEventLoop(virNetClientPtr client, } if (ret < 0) { + /* XXX what's this dubious errno check doing ? */ if (errno == EWOULDBLOCK) continue; virReportSystemError(errno, @@ -1005,8 +1068,8 @@ static int virNetClientIOEventLoop(virNetClientPtr client, goto error; } - /* Iterate through waiting threads and if - * any are complete then tell 'em to wakeup + /* Iterate through waiting calls and if any are + * complete, remove them from the dispatch list.. */ tmp = client->waitDispatch; prev = NULL; @@ -1019,13 +1082,25 @@ static int virNetClientIOEventLoop(virNetClientPtr client, else client->waitDispatch = tmp->next; - /* And wake them up.... - * ...they won't actually wakeup until + /* + * ...if the call being removed from the list + * still has a thread, then wake that thread up, + * otherwise free the call. The latter should + * only happen for calls without replies. + * + * ...the threads won't actually wakeup until * we release our mutex a short while * later... */ - VIR_DEBUG("Waking up sleep %p %p", tmp, client->waitDispatch); - virCondSignal(&tmp->cond); + if (tmp->haveThread) { + VIR_DEBUG("Waking up sleep %p %p", tmp, client->waitDispatch); + virCondSignal(&tmp->cond); + } else { + if (tmp->expectReply) + VIR_WARN("Got a call expecting a reply but without a waiting thread"); + ignore_value(virCondDestroy(&tmp->cond)); + VIR_FREE(tmp); + } } else { prev = tmp; } @@ -1039,13 +1114,8 @@ static int virNetClientIOEventLoop(virNetClientPtr client, */ client->waitDispatch = thiscall->next; VIR_DEBUG("Giving up the buck %p %p", thiscall, client->waitDispatch); - /* See if someone else is still waiting - * and if so, then pass the buck ! */ - if (client->waitDispatch) { - VIR_DEBUG("Passing the buck to %p", client->waitDispatch); - virCondSignal(&client->waitDispatch->cond); - } - return 0; + virNetClientPassTheBuck(client); + return 2; } @@ -1060,16 +1130,21 @@ static int virNetClientIOEventLoop(virNetClientPtr client, error: client->waitDispatch = thiscall->next; VIR_DEBUG("Giving up the buck due to I/O error %p %p", thiscall, client->waitDispatch); - /* See if someone else is still waiting - * and if so, then pass the buck ! */ - if (client->waitDispatch) { - VIR_DEBUG("Passing the buck to %p", client->waitDispatch); - virCondSignal(&client->waitDispatch->cond); - } + virNetClientPassTheBuck(client); return -1; } +static void +virNetClientUpdateIOCallback(virNetClientPtr client, bool enabled) +{ + int events = 0; + if (enabled) { + events |= VIR_EVENT_HANDLE_READABLE; + } + virNetSocketUpdateIOCallback(client->sock, events); +} + /* * This function sends a message to remote server and awaits a reply * @@ -1102,11 +1177,15 @@ error: * nation are blamed on another, providing an opportunity for war." * * NB(5) Don't Panic! + * + * Returns 2 if fully sent, 1 if partially sent (only for nonBlock==true), + * 0 if nothing sent (only for nonBlock==true) and -1 on error */ static int virNetClientIO(virNetClientPtr client, virNetClientCallPtr thiscall) { int rv = -1; + virNetClientCallPtr tmp; VIR_DEBUG("Outgoing message prog=%u version=%u serial=%u proc=%d type=%d length=%zu dispatch=%p", thiscall->msg->header.prog, @@ -1117,20 +1196,27 @@ static int virNetClientIO(virNetClientPtr client, thiscall->msg->bufferLength, client->waitDispatch); + /* Trivially detect blocking if someone else has the buck already */ + if (client->haveTheBuck && + thiscall->nonBlock) + return 0; + + /* Stick ourselves on the end of the wait queue */ + tmp = client->waitDispatch; + while (tmp && tmp->next) + tmp = tmp->next; + if (tmp) + tmp->next = thiscall; + else + client->waitDispatch = thiscall; + /* Check to see if another thread is dispatching */ - if (client->waitDispatch) { - /* Stick ourselves on the end of the wait queue */ - virNetClientCallPtr tmp = client->waitDispatch; + if (client->haveTheBuck) { char ignore = 1; - while (tmp && tmp->next) - tmp = tmp->next; - if (tmp) - tmp->next = thiscall; - else - client->waitDispatch = thiscall; /* Force other thread to wakeup from poll */ if (safewrite(client->wakeupSendFD, &ignore, sizeof(ignore)) != sizeof(ignore)) { + /* Something went wrong, so we need to remove that call we just added */ if (tmp) tmp->next = NULL; else @@ -1143,6 +1229,7 @@ static int virNetClientIO(virNetClientPtr client, VIR_DEBUG("Going to sleep %p %p", client->waitDispatch, thiscall); /* Go to sleep while other thread is working... */ if (virCondWait(&thiscall->cond, &client->lock) < 0) { + /* Something went wrong, so we need to remove that call we previously added */ if (client->waitDispatch == thiscall) { client->waitDispatch = thiscall->next; } else { @@ -1167,7 +1254,7 @@ static int virNetClientIO(virNetClientPtr client, * our reply */ if (thiscall->mode == VIR_NET_CLIENT_MODE_COMPLETE) { - rv = 0; + rv = 2; /* * We avoided catching the buck and our reply is ready ! * We've already had 'thiscall' removed from the list @@ -1177,12 +1264,10 @@ static int virNetClientIO(virNetClientPtr client, } /* Grr, someone passed the buck onto us ... */ - - } else { - /* We're first to catch the buck */ - client->waitDispatch = thiscall; } + client->haveTheBuck = true; + VIR_DEBUG("We have the buck %p %p", client->waitDispatch, thiscall); /* * The buck stops here! @@ -1198,17 +1283,19 @@ static int virNetClientIO(virNetClientPtr client, * cause the event loop thread to be blocked on the * mutex for the duration of the call */ - virNetSocketUpdateIOCallback(client->sock, 0); + virNetClientUpdateIOCallback(client, false); virResetLastError(); rv = virNetClientIOEventLoop(client, thiscall); - virNetSocketUpdateIOCallback(client->sock, VIR_EVENT_HANDLE_READABLE); + virNetClientUpdateIOCallback(client, true); if (rv == 0 && virGetLastError()) rv = -1; + client->haveTheBuck = false; + cleanup: VIR_DEBUG("All done with our call %p %p %d", client->waitDispatch, thiscall, rv); return rv; @@ -1227,7 +1314,7 @@ void virNetClientIncomingEvent(virNetSocketPtr sock, goto done; /* This should be impossible, but it doesn't hurt to check */ - if (client->waitDispatch) + if (client->haveTheBuck) goto done; VIR_DEBUG("Event fired %p %d", sock, events); @@ -1249,9 +1336,14 @@ done: } -int virNetClientSend(virNetClientPtr client, - virNetMessagePtr msg, - bool expectReply) +/* + * Returns 2 if fully sent, 1 if partially sent (only for nonBlock==true), + * 0 if nothing sent (only for nonBlock==true) and -1 on error + */ +static int virNetClientSendInternal(virNetClientPtr client, + virNetMessagePtr msg, + bool expectReply, + bool nonBlock) { virNetClientCallPtr call; int ret = -1; @@ -1270,6 +1362,12 @@ int virNetClientSend(virNetClientPtr client, return -1; } + if (expectReply && nonBlock) { + virNetError(VIR_ERR_INTERNAL_ERROR, "%s", + _("Attempt to send an non-blocking message with a synchronous reply")); + return -1; + } + if (VIR_ALLOC(call) < 0) { virReportOOMError(); return -1; @@ -1290,12 +1388,75 @@ int virNetClientSend(virNetClientPtr client, call->mode = VIR_NET_CLIENT_MODE_WAIT_RX; call->msg = msg; call->expectReply = expectReply; + call->nonBlock = nonBlock; + call->haveThread = true; ret = virNetClientIO(client, call); cleanup: - ignore_value(virCondDestroy(&call->cond)); - VIR_FREE(call); + /* If partially sent, then the call is still on the dispatch queue */ + if (ret == 1) { + call->haveThread = false; + } else { + ignore_value(virCondDestroy(&call->cond)); + VIR_FREE(call); + } virNetClientUnlock(client); return ret; } + + +/* + * @msg: a message allocated on heap or stack + * + * Send a message synchronously, and wait for the reply synchronously + * + * The caller is responsible for free'ing @msg if it was allocated + * on the heap + * + * Returns 0 on success, -1 on failure + */ +int virNetClientSend(virNetClientPtr client, + virNetMessagePtr msg) +{ + int ret = virNetClientSendInternal(client, msg, true, false); + if (ret < 0) + return -1; + return 0; +} + + +/* + * @msg: a message allocated on heap or stack + * + * Send a message synchronously, without any reply + * + * The caller is responsible for free'ing @msg if it was allocated + * on the heap + * + * Returns 0 on success, -1 on failure + */ +int virNetClientSendNoReply(virNetClientPtr client, + virNetMessagePtr msg) +{ + int ret = virNetClientSendInternal(client, msg, false, false); + if (ret < 0) + return -1; + return 0; +} + +/* + * @msg: a message allocated on the heap. + * + * Send a message asynchronously, without any reply + * + * The caller is responsible for free'ing @msg, *except* if + * this method returns -1. + * + * Returns 2 on full send, 1 on partial send, 0 on no send, -1 on error + */ +int virNetClientSendNonBlock(virNetClientPtr client, + virNetMessagePtr msg) +{ + return virNetClientSendInternal(client, msg, false, true); +} diff --git a/src/rpc/virnetclient.h b/src/rpc/virnetclient.h index fb679e8..d3c112a 100644 --- a/src/rpc/virnetclient.h +++ b/src/rpc/virnetclient.h @@ -67,9 +67,17 @@ int virNetClientAddStream(virNetClientPtr client, void virNetClientRemoveStream(virNetClientPtr client, virNetClientStreamPtr st); +/* Send a message and wait for reply */ int virNetClientSend(virNetClientPtr client, - virNetMessagePtr msg, - bool expectReply); + virNetMessagePtr msg); + +/* Send a message without needing a reply */ +int virNetClientSendNoReply(virNetClientPtr client, + virNetMessagePtr msg); + +/* Send a message without needing a reply, and don't block on I/O */ +int virNetClientSendNonBlock(virNetClientPtr client, + virNetMessagePtr msg); # ifdef HAVE_SASL void virNetClientSetSASLSession(virNetClientPtr client, diff --git a/src/rpc/virnetclientprogram.c b/src/rpc/virnetclientprogram.c index 36e2384..cb02a25 100644 --- a/src/rpc/virnetclientprogram.c +++ b/src/rpc/virnetclientprogram.c @@ -327,7 +327,7 @@ int virNetClientProgramCall(virNetClientProgramPtr prog, if (virNetMessageEncodePayload(msg, args_filter, args) < 0) goto error; - if (virNetClientSend(client, msg, true) < 0) + if (virNetClientSend(client, msg) < 0) goto error; /* None of these 3 should ever happen here, because diff --git a/src/rpc/virnetclientstream.c b/src/rpc/virnetclientstream.c index 7e2d9ae..309d48d 100644 --- a/src/rpc/virnetclientstream.c +++ b/src/rpc/virnetclientstream.c @@ -361,8 +361,13 @@ int virNetClientStreamSendPacket(virNetClientStreamPtr st, wantReply = true; } - if (virNetClientSend(client, msg, wantReply) < 0) - goto error; + if (wantReply) { + if (virNetClientSend(client, msg) < 0) + goto error; + } else { + if (virNetClientSendNoReply(client, msg) < 0) + goto error; + } virNetMessageFree(msg); @@ -407,7 +412,7 @@ int virNetClientStreamRecvPacket(virNetClientStreamPtr st, VIR_DEBUG("Dummy packet to wait for stream data"); virMutexUnlock(&st->lock); - ret = virNetClientSend(client, msg, true); + ret = virNetClientSend(client, msg); virMutexLock(&st->lock); virNetMessageFree(msg); -- 1.7.6.4

2 1

[libvirt] [PATCH] Don't return a fatal error if receiving unexpected stream data
by Daniel P. Berrange 09 Nov '11

09 Nov '11

From: "Daniel P. Berrange" <berrange(a)redhat.com> Due to the asynchronous nature of streams, we might continue to receive some stream packets from the server even after we have shutdown the stream on the client side. These should be discarded silently, rather than raising an error in the RPC layer. * src/rpc/virnetclient.c: Discard stream data silently --- src/rpc/virnetclient.c | 4 +++- 1 files changed, 3 insertions(+), 1 deletions(-) diff --git a/src/rpc/virnetclient.c b/src/rpc/virnetclient.c index 4b7d4a9..4cbdf55 100644 --- a/src/rpc/virnetclient.c +++ b/src/rpc/virnetclient.c @@ -621,7 +621,9 @@ static int virNetClientCallDispatchStream(virNetClientPtr client) VIR_DEBUG("No stream found for packet with prog=%d vers=%d serial=%u proc=%u", client->msg.header.prog, client->msg.header.vers, client->msg.header.serial, client->msg.header.proc); - return -1; + /* Don't return -1, because we expect to see further stream packets + * after we've shut it down sometimes */ + return 0; } /* Finish/Abort are synchronous, so also see if there's an -- 1.7.6.4

2 1

[libvirt] ANNOUNCE: Release 0.9.7 of Perl Bindings Sys::Virt
by Daniel P. Berrange 09 Nov '11

09 Nov '11

I am pleased to announce an update of the Perl bindings Sys::Virt, version 0.9.7 is now available for download: http://search.cpan.org/CPAN/authors/id/D/DA/DANBERR/Sys-Virt-0.9.7.tar.gz New in 0.9.7: - Add all new APIs in libvirt 0.9.7 - Add binding for virDomainOpenConsole - Fix missing $flags in documentation - Added example integrating Sys::Virt::Domain->open_graphics GTK3 and GTK-VNC The permalink for Sys-Virt, where all previous releases can also be obtained, is: http://search.cpan.org/dist/Sys-Virt/ Regards, Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|

1 0

[libvirt] [PATCH] lxc: plug memory leak
by ajia＠redhat.com 09 Nov '11

09 Nov '11

From: Alex Jia <ajia(a)redhat.com> Detected by Coverity. Leak introduced in commit 0f31f7b. * src/lxc/lxc_driver.c: Clean up on failure. Signed-off-by: Alex Jia <ajia(a)redhat.com> --- src/lxc/lxc_driver.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/src/lxc/lxc_driver.c b/src/lxc/lxc_driver.c index 37092bc..a1d0a7c 100644 --- a/src/lxc/lxc_driver.c +++ b/src/lxc/lxc_driver.c @@ -1850,6 +1850,7 @@ cleanup: } for (i = 0 ; i < nttyFDs ; i++) VIR_FORCE_CLOSE(ttyFDs[i]); + VIR_FREE(ttyFDs); VIR_FORCE_CLOSE(handshakefds[0]); VIR_FORCE_CLOSE(handshakefds[1]); VIR_FREE(logfile); -- 1.7.1

3 2

[libvirt] [PATCH] lxc: plug memory leak
by ajia＠redhat.com 09 Nov '11

09 Nov '11

From: Alex Jia <ajia(a)redhat.com> Detected by Coverity. Leak introduced in commit 9d201a5. * src/lxc/lxc_driver.c: Clean up on failure. Signed-off-by: Alex Jia <ajia(a)redhat.com> --- src/lxc/lxc_driver.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/lxc/lxc_driver.c b/src/lxc/lxc_driver.c index 37092bc..295fd6f 100644 --- a/src/lxc/lxc_driver.c +++ b/src/lxc/lxc_driver.c @@ -1856,7 +1856,7 @@ cleanup: if (err) { virSetError(err); - virResetError(err); + virFreeError(err); } return rc; -- 1.7.1

2 1

[libvirt] [PATCH] qemu: Fix improper error message for disk detaching
by Osier Yang 09 Nov '11

09 Nov '11

s/virDomainDeviceTypeToString/virDomainDiskDeviceTypeToString/ Report by Xu He Jie <xuhj(a)linux.vnet.ibm.com> --- src/qemu/qemu_driver.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 118197a..cee689b 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -5212,8 +5212,8 @@ qemuDomainDetachDeviceDiskLive(struct qemud_driver *driver, break; default: qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED, - _("device type '%s' cannot be detached"), - virDomainDeviceTypeToString(dev->type)); + _("disk device type '%s' cannot be detached"), + virDomainDiskDeviceTypeToString(dev->type)); break; } return ret; -- 1.7.6

2 2

[libvirt] [PATCH] virsh: Add VSH_OFLAG_EMPTY_OK for attach-disk command
by Xu He Jie 09 Nov '11

09 Nov '11

As the description of removing CDROM media from http://wiki.libvirt.org/page/QEMUSwitchToLibvirt#eject_DEV Add flag 'VSH_OFLAG_EMPTY_OK' to the option 'source' of attach-disk Signed-off-by: Xu He Jie <xuhj(a)linux.vnet.ibm.com> --- tools/virsh.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/tools/virsh.c b/tools/virsh.c index 5544a41..e7eae74 100644 --- a/tools/virsh.c +++ b/tools/virsh.c @@ -11569,7 +11569,7 @@ static const vshCmdInfo info_attach_disk[] = { static const vshCmdOptDef opts_attach_disk[] = { {"domain", VSH_OT_DATA, VSH_OFLAG_REQ, N_("domain name, id or uuid")}, - {"source", VSH_OT_DATA, VSH_OFLAG_REQ, N_("source of disk device")}, + {"source", VSH_OT_DATA, VSH_OFLAG_REQ | VSH_OFLAG_EMPTY_OK, N_("source of disk device")}, {"target", VSH_OT_DATA, VSH_OFLAG_REQ, N_("target of disk device")}, {"driver", VSH_OT_STRING, 0, N_("driver of disk device")}, {"subdriver", VSH_OT_STRING, 0, N_("subdriver of disk device")}, -- 1.7.5.4

3 5

[libvirt] [PATCHv2] build: fix build on platforms without ptsname_r
by Eric Blake 08 Nov '11

08 Nov '11

MacOS lacks ptsname_r, and gnulib doesn't (yet) provide it. But we can avoid it altogether, by using gnulib openpty() instead. Note that we do _not_ want the pt_chown module; all systems that we currently port to can either properly do openpty() and/or grantpt(), or lack ptys altogether; we are not porting to any system that requires us to deal with the hassle of installing a setuid pt_chown helper just to satisfy gnulib's ability to provide openpty() on even more platforms. * .gnulib: Update to latest, for openpty fixes * bootstrap.conf (gnulib_modules): Add openpty, ttyname_r. (gnulib_tool_option_extras): Exclude pt_chown module. * src/util/util.c (virFileOpenTty): Rewrite in terms of openpty and ttyname_r. * src/util/util.h (virFileOpenTtyAt): Delete dead prototype. --- Alas, this is just complicated enough that I don't feel comfortable pushing it under the build-breaker rule, even though I have verified that it fixes builds on both FreeBSD and Cygwin, as well as still behaves correctly with LXC containers on Linux. Anyone willing to give this a review before 0.9.7 gets released? .gnulib | 2 +- bootstrap.conf | 3 ++ src/util/util.c | 73 ++++++++++++++++++++++++++++++++++++++++-------------- src/util/util.h | 5 ---- 4 files changed, 58 insertions(+), 25 deletions(-) diff --git a/.gnulib b/.gnulib index 2394a60..0031e4f 160000 --- a/.gnulib +++ b/.gnulib @@ -1 +1 @@ -Subproject commit 2394a603e7586e671226478e5b15d924c3841f42 +Subproject commit 0031e4f6353cc7077a9d0dad0c793bd6e3dc7aaa diff --git a/bootstrap.conf b/bootstrap.conf index 0faa2e2..4557d2d 100644 --- a/bootstrap.conf +++ b/bootstrap.conf @@ -68,6 +68,7 @@ mkstemps mktempd netdb nonblocking +openpty passfd perror physmem @@ -101,6 +102,7 @@ sys_wait termios time_r timegm +ttyname_r uname useless-if-before-free usleep @@ -168,6 +170,7 @@ tests_base=gnulib/tests gnulib_tool_option_extras="\ --lgpl=2\ --with-tests\ + --avoid=pt_chown\ " # Convince bootstrap to use multiple m4 directories. diff --git a/src/util/util.c b/src/util/util.c index bd52b21..959c224 100644 --- a/src/util/util.c +++ b/src/util/util.c @@ -45,10 +45,11 @@ #include <string.h> #include <signal.h> #include <termios.h> +#include <pty.h> + #if HAVE_LIBDEVMAPPER_H # include <libdevmapper.h> #endif -#include "c-ctype.h" #ifdef HAVE_PATHS_H # include <paths.h> @@ -65,6 +66,7 @@ # include <mntent.h> #endif +#include "c-ctype.h" #include "dirname.h" #include "virterror_internal.h" #include "logging.h" @@ -1172,52 +1174,85 @@ virFileBuildPath(const char *dir, const char *name, const char *ext) return path; } +/* Open a non-blocking master side of a pty. If ttyName is not NULL, + * then populate it with the name of the slave. If rawmode is set, + * also put the master side into raw mode before returning. */ #ifndef WIN32 int virFileOpenTty(int *ttymaster, char **ttyName, int rawmode) { - int rc = -1; - - if ((*ttymaster = posix_openpt(O_RDWR|O_NOCTTY|O_NONBLOCK)) < 0) - goto cleanup; + /* XXX A word of caution - on some platforms (Solaris and HP-UX), + * additional ioctl() calls are needs after opening the slave + * before it will cause isatty() to return true. Should we make + * virFileOpenTty also return the opened slave fd, so the caller + * doesn't have to worry about that mess? */ + int ret = -1; + int slave = -1; + char *name = NULL; - if (unlockpt(*ttymaster) < 0) - goto cleanup; + /* Unfortunately, we can't use the name argument of openpty, since + * there is no guarantee on how large the buffer has to be. + * Likewise, we can't use the termios argument: we have to use + * read-modify-write since there is no portable way to initialize + * a struct termios without use of tcgetattr. */ + if (openpty(ttymaster, &slave, NULL, NULL, NULL) < 0) + return -1; - if (grantpt(*ttymaster) < 0) + /* What a shame that openpty cannot atomically set FD_CLOEXEC, but + * that using posix_openpt/grantpt/unlockpt/ptsname is not + * thread-safe, and that ptsname_r is not portable. */ + if (virSetNonBlock(*ttymaster) < 0 || + virSetCloseExec(*ttymaster) < 0) goto cleanup; + /* While Linux supports tcgetattr on either the master or the + * slave, Solaris requires it to be on the slave. */ if (rawmode) { struct termios ttyAttr; - if (tcgetattr(*ttymaster, &ttyAttr) < 0) + if (tcgetattr(slave, &ttyAttr) < 0) goto cleanup; cfmakeraw(&ttyAttr); - if (tcsetattr(*ttymaster, TCSADRAIN, &ttyAttr) < 0) + if (tcsetattr(slave, TCSADRAIN, &ttyAttr) < 0) goto cleanup; } + /* ttyname_r on the slave is required by POSIX, while ptsname_r on + * the master is a glibc extension, and the POSIX ptsname is not + * thread-safe. Since openpty gave us both descriptors, guess + * which way we will determine the name? :) */ if (ttyName) { - if (VIR_ALLOC_N(*ttyName, PATH_MAX) < 0) { - errno = ENOMEM; + /* Initial guess of 64 is generally sufficient; rely on ERANGE + * to tell us if we need to grow. */ + size_t len = 64; + int rc; + + if (VIR_ALLOC_N(name, len) < 0) goto cleanup; - } - if (ptsname_r(*ttymaster, *ttyName, PATH_MAX) != 0) { - VIR_FREE(*ttyName); + while ((rc = ttyname_r(slave, name, len)) == ERANGE) { + if (VIR_RESIZE_N(name, len, len, len) < 0) + goto cleanup; + } + if (rc != 0) { + errno = rc; goto cleanup; } + *ttyName = name; + name = NULL; } - rc = 0; + ret = 0; cleanup: - if (rc != 0) + if (ret != 0) VIR_FORCE_CLOSE(*ttymaster); + VIR_FORCE_CLOSE(slave); + VIR_FREE(name); - return rc; + return ret; } #else /* WIN32 */ int virFileOpenTty(int *ttymaster ATTRIBUTE_UNUSED, diff --git a/src/util/util.h b/src/util/util.h index e869f1b..d8176a8 100644 --- a/src/util/util.h +++ b/src/util/util.h @@ -121,11 +121,6 @@ int virFileAbsPath(const char *path, int virFileOpenTty(int *ttymaster, char **ttyName, int rawmode); -int virFileOpenTtyAt(const char *ptmx, - int *ttymaster, - char **ttyName, - int rawmode); - char *virArgvToString(const char *const *argv); -- 1.7.4.4

2 4

[libvirt] [libvirt-glib 00/24] libvirt-gconfig patches
by Christophe Fergeau 08 Nov '11

08 Nov '11

Hi, Here is a set of patches which start to implement functionality to libvirt-gconfig. The end result is that the "name", "memory" and "features" properties are implemented for the Domain class. Most of the work is about reworking the low-level plumbing to make the implementation as easy as possible. I chose to remove the xmlDocPtr and the xml string that were stored in GVirConfigObject objects and to replace this with a xmlNodePtr. This is useful when creating children objects (eg devices) when parsing a domain XML because the xmlNodePtr will point at the right position in the XML document. One caveat with that approach is that we need to refcount the xmlDocPtr associated with the xmlNodePtr to know when to free it, but this is not implemented yet (though gupnp seems to have code doing exactly this). Then most of the work is adding properties and helpers to make the code easier to write. And thinking about what API we want to expose to clients :) The two test programs should show how the API can be used as it is now. While writing this, I realized we had to do some "useless" work to return strings from libxml2 to glib, ie I g_strdup the string and xmlFree the source so that the caller can free it with g_free. While writing this code, there were at least 2 questions that arose, some input would be welcome :) * I don't know how to handle XMLs where the same node appears multiple times (eg "name"). If it's the latest name that wins, what do we do when changing the name on such an XML document? only modify the last name node? Drop all the redundant name nodes? * I'm also not sure how to make it possible to check whether a given property is set or not. For strings, this is easy, NULL means that the property was not set, but for integers, this is less obvious. Any thoughts on that? Next I'll have to start looking at more complicated properties :) More API questions on the way I guess... Christophe Christophe Fergeau (23): Add helpers in libvirt-gconfig-helpers.[ch] gvir_config_object_parse: don't parse empty documents Add getters for GVirConfig xmlNode and xmlDoc Add GVirConfigObject::node property Add GVir::Config::Domain::name property Rename gvir_config_domain_new Add gvir_config_domain_new to create an empty domain Implement gvir_config_domain_set_name Make the GVirConfigDomain::name property writable Add gvir_config_object_to_xml Use gvir_config_object_to_xml Add domain creation/parsing test Remove GVirConfigObject::docHandle Remove xml parsing from gvir_config_*_new functions Only do XML parsing when creating config objects Remove GError argument from GVirConfigObject::node getter Remove GVirConfigObject::doc Add gvir_config_object_get_node_content Add gvir_config_object_set_node_content Add test-domain-duplicate.xml which currently fails Validate document in parsing test Add GVirConfigDomain::memory Add GVirConfigDomain::features configure.ac | 1 + examples/Makefile.am | 1 + libvirt-gconfig/Makefile.am | 4 + libvirt-gconfig/libvirt-gconfig-capabilities.c | 10 +- libvirt-gconfig/libvirt-gconfig-capabilities.h | 2 +- libvirt-gconfig/libvirt-gconfig-domain-snapshot.c | 9 +- libvirt-gconfig/libvirt-gconfig-domain-snapshot.h | 2 +- libvirt-gconfig/libvirt-gconfig-domain.c | 186 ++++++++++++++++- libvirt-gconfig/libvirt-gconfig-domain.h | 12 +- libvirt-gconfig/libvirt-gconfig-helpers.c | 179 +++++++++++++++ libvirt-gconfig/libvirt-gconfig-helpers.h | 48 ++++ libvirt-gconfig/libvirt-gconfig-interface.c | 12 +- libvirt-gconfig/libvirt-gconfig-interface.h | 2 +- libvirt-gconfig/libvirt-gconfig-network-filter.c | 8 +- libvirt-gconfig/libvirt-gconfig-network-filter.h | 2 +- libvirt-gconfig/libvirt-gconfig-network.c | 8 +- libvirt-gconfig/libvirt-gconfig-network.h | 2 +- libvirt-gconfig/libvirt-gconfig-node-device.c | 9 +- libvirt-gconfig/libvirt-gconfig-node-device.h | 2 +- libvirt-gconfig/libvirt-gconfig-object.c | 244 ++++++++++++--------- libvirt-gconfig/libvirt-gconfig-object.h | 17 ++- libvirt-gconfig/libvirt-gconfig-secret.c | 8 +- libvirt-gconfig/libvirt-gconfig-secret.h | 2 +- libvirt-gconfig/libvirt-gconfig-storage-pool.c | 8 +- libvirt-gconfig/libvirt-gconfig-storage-pool.h | 2 +- libvirt-gconfig/libvirt-gconfig-storage-vol.c | 12 +- libvirt-gconfig/libvirt-gconfig-storage-vol.h | 2 +- libvirt-gconfig/libvirt-gconfig.h | 2 + libvirt-gconfig/libvirt-gconfig.sym | 11 +- libvirt-gconfig/tests/Makefile.am | 16 ++ libvirt-gconfig/tests/test-domain-create.c | 68 ++++++ libvirt-gconfig/tests/test-domain-duplicate.xml | 7 + libvirt-gconfig/tests/test-domain-invalid.xml | 6 + libvirt-gconfig/tests/test-domain-noname.xml | 4 + libvirt-gconfig/tests/test-domain-parse.c | 87 ++++++++ libvirt-gconfig/tests/test-domain.xml | 7 + libvirt-gobject/Makefile.am | 2 + libvirt-gobject/libvirt-gobject-connection.c | 2 +- libvirt-gobject/libvirt-gobject-domain-snapshot.c | 3 + libvirt-gobject/libvirt-gobject-domain.c | 6 +- libvirt-gobject/libvirt-gobject-interface.c | 3 + libvirt-gobject/libvirt-gobject-network-filter.c | 3 + libvirt-gobject/libvirt-gobject-network.c | 3 + libvirt-gobject/libvirt-gobject-node-device.c | 4 + libvirt-gobject/libvirt-gobject-secret.c | 4 + libvirt-gobject/libvirt-gobject-storage-pool.c | 5 +- libvirt-gobject/libvirt-gobject-storage-vol.c | 3 + 47 files changed, 895 insertions(+), 145 deletions(-) create mode 100644 libvirt-gconfig/libvirt-gconfig-helpers.c create mode 100644 libvirt-gconfig/libvirt-gconfig-helpers.h create mode 100644 libvirt-gconfig/tests/Makefile.am create mode 100644 libvirt-gconfig/tests/test-domain-create.c create mode 100644 libvirt-gconfig/tests/test-domain-duplicate.xml create mode 100644 libvirt-gconfig/tests/test-domain-invalid.xml create mode 100644 libvirt-gconfig/tests/test-domain-noname.xml create mode 100644 libvirt-gconfig/tests/test-domain-parse.c create mode 100644 libvirt-gconfig/tests/test-domain.xml -- 1.7.6.4

3 62

[libvirt] qemu error: Could not allocate dynamic translator buffer
by Richard W.M. Jones 08 Nov '11

08 Nov '11

I'm getting a very odd error from qemu: 2011-11-08 13:06:03.403: starting up LC_ALL=C PATH=/sbin:/usr/sbin:/bin:/usr/bin QEMU_AUDIO_DRV=none /usr/bin/qemu-system-x86_64 -S -M pc-0.14 -no-kvm -m 1024 -smp 1,sockets=1,cores=1,threads=1 -name fedora15_x86_64 -uuid a88161b0-c4c9-42bf-a7b5-5bccf8b65197 -nodefconfig -nodefaults -chardev socket,id=charmonitor,path=/var/lib/libvirt/qemu/fedora15_x86_64.monitor,server,nowait -mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc -no-reboot -kernel /var/lib/libvirt/images/fedora15_x86_64-kernel -initrd /var/lib/libvirt/images/fedora15_x86_64-ramdisk -append method=http://mirror.bytemark.co.uk/fedora/linux/releases/15/Fedora/x86_64/os/ ks=file:/ks.cfg -drive file=/var/lib/libvirt/images/fedora15_x86_64.dsk,if=none,id=drive-virtio-disk0,format=raw -device virtio-blk-pci,bus=pci.0,addr=0x4,drive=drive-virtio-disk0,id=virtio-disk0 -netdev tap,fd=24,id=hostnet0 -device virtio-net-pci,netdev=hostnet0,id=net0,mac=52:54:00:26:9f:d6,bus=pci.0,addr=0x3 -chardev socket,id=charserial0,host=127.0.0.1,port=64900,server,nowait -device isa-serial,chardev=charserial0,id=serial0 -usb -vnc 127.0.0.1:0 -vga cirrus -device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x5 Could not allocate dynamic translator buffer 2011-11-08 13:06:04.062: shutting down Now I understand that qemu gives this error when it cannot allocate memory, specifically it can't mmap a TCG translation buffer. 800 MBytes according to the source code. However the guest where I'm running this command has plenty of free memory available: $ free -m total used free shared buffers cached Mem: 5659 872 4787 0 26 351 -/+ buffers/cache: 493 5166 Swap: 1983 0 1983 That's surely easily enough for 1G of RAM + a translation buffer? Anyway, my question is this: Can I somehow get libvirt to strace qemu and put that output somewhere? I tried the strace trick here: https://fedoraproject.org/wiki/How_to_debug_Virtualization_problems#strace but for some reason it doesn't trace into the qemu process. Not sure why, since surely the qemu process is a child of libvirtd? Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming blog: http://rwmj.wordpress.com Fedora now supports 80 OCaml packages (the OPEN alternative to F#) http://cocan.org/getting_started_with_ocaml_on_red_hat_and_fedora

1 1

[libvirt] build failed with warning " failed to load external entity "api.html.tmp"
by Wenyi Gao 08 Nov '11

08 Nov '11

I got the latest code and build it with following error on Ubuntu 11.10: ...... make all-am make[3]: Entering directory `/home/wayne/workspace/project/libvirt/tools' CC virsh-console.o CC virsh-virsh.o CCLD virsh GEN virt-xml-validate GEN virt-pki-validate make[3]: Leaving directory `/home/wayne/workspace/project/libvirt/tools' make[2]: Leaving directory `/home/wayne/workspace/project/libvirt/tools' Making all in docs make[2]: Entering directory `/home/wayne/workspace/project/libvirt/docs' Validating api.html warning: failed to load external entity "api.html.tmp" make[2]: *** [api.html] Error 1 make[2]: Leaving directory `/home/wayne/workspace/project/libvirt/docs' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory `/home/wayne/workspace/project/libvirt' make: *** [all] Error 2 I think some dependent packages are missed on my build environment. But the error information doesn't show some hint. Does anyone know the reason? Thanks. Wenyi Gao

2 1

[libvirt] Release of libvirt-0.9.7
by Daniel Veillard 08 Nov '11

08 Nov '11

With a bit of delay to catch the last set of patches, the release is finally out ! You can get 0.9.7 from ftp://libvirt.org/libvirt/ This is a release mostly focusing on bug fixes and cleanup though there is some features too, including a lot of improvements on the snapshot code. There is also a notable set of portability patches, hopefully all issues raised there were fixed ! Features: - esx: Support vSphere 5.x (Patrice LACHANCE) - vbox: Add support for VirtualBox 4.1 (Matthias Bolte) - Introduce the virDomainOpenGraphics API (Daniel P. Berrange) - Add AHCI support to qemu driver (Jim Fehlig) - snapshot: many improvements and 2 new APIs (Eric Blake) - api: Add public api for 'reset' (Xu He Jie) Documentation: - Add <deviceboot> capability. (Philipp Hahn) - API: document scheduler parameter names (Eric Blake) - improve typed parameter documentation (Eric Blake) - fix typo in <disk>/<target> example (Eric Blake) - python: Fix documentation of virStream recv (Matthias Bolte) - Add documentation about migration. (Daniel P. Berrange) - Document filesystem type='block' for LXC (Ryota Ozaki) - fix incorrect info about routed networks (Laine Stump) - document managed=yes of hostdev passthrough (Eric Blake) - trivial spelling fix (Philipp Hahn) - Fix virt-sanlock-cleanup documentation (Philipp Hahn) - Fix two comments related to error handling (Philipp Hahn) - fix network XML documentation (Laine Stump) - Clarify semantics of virDomainMigrate2 (Jiri Denemark) - Update examples for probing with systemtap (Daniel P. Berrange) - fix html bug (Eric Blake) - virsh: Update the help information for undefine command. (tangchen) - Document STREQ_NULLABLE and STRNEQ_NULLABLE (Guido Günther) - Document that ff callbacks need to be invoked from a clean stack. (Guido Günther) - formatdomain.html.in: fix tickpolicy (Douglas Schilling Landgraf) - virsh: describe attach-interface parameter target (Daniel Veillard) - virsh: update man page for cpu_shares parameter (Daniel Veillard) - document node device XML (Eric Blake) - document virsh nodedev-* commands (Eric Blake) - snapshot: fix man page typos (Eric Blake) - virsh: Better document --copy-storage migrate options (Jiri Denemark) - virsh: Enhance documentation of commands starting jobs (Jiri Denemark) - virsh: Improve virsh manual for virsh memtune command (Peter Krempa) Portability: - build: fix build on platforms without ptsname_r (Eric Blake) - build: silence compiler warning on BSD (Eric Blake) - build: fix linking on BSD (Eric Blake) - remote: fix mingw32 build (Laine Stump) - build: fix deep VPATH builds (Eric Blake) - Use ENAMETOOLONG if the the socket path is longer than UNIX_PATH_MAX (Guido Günther) - build: avoid RHEL 5 build failure on LXC (Eric Blake) - build: use gnulib fdatasync (Eric Blake) - Fix virFileOpenTty definition on Win32 (Daniel P. Berrange) - compile: fix undefined reference to gnutls_x509_crt_get_dn with gcc-4.6.1 (Xu He Jie) - Fix VPATH build (Jiri Denemark) - build: fix 'make dist' error (Wen Congyang) - Fix syntax problem in mingw32-libvirt.spec.in (Daniel P. Berrange) - spec: mingw cleanups (Eric Blake) - build: fix mingw build without sasl (Eric Blake) - build: fix 'make rpm' (Eric Blake) - build: fix 'make distcheck' (Eric Blake) - disable xenlight for non-Xen platforms (Dan Horák) - build: Fix VPATH build with new probes (Jiri Denemark) - build: fix 'make distcheck' with pdwtags installed (Eric Blake) - spec: F15 still uses cgconfig, RHEL lacks hyperv (Eric Blake) Bug Fixes: - Fix sending/receiving of FDs when stream returns EAGAIN (Daniel P. Berrange) - lxc: avoid use-after-free (Eric Blake) - conf: Don't free uninitialized pointer (Jiri Denemark) - Fix default console type setting (Daniel P. Berrange) - Fix crash formatting virtio console (Daniel P. Berrange) - Fix off-by-one printing month in logging code (Daniel P. Berrange) - Add missing param initialization in qemuDomainBlockStatsFlags (Daniel P. Berrange) - fix crash when starting network (Wen Congyang) - Don't overwrite error message during VM cleanup (Daniel P. Berrange) - Correctly handle '*' in /etc/filesystems (Daniel P. Berrange) - Fix URI alias prefix matching (Wen Ruo Lv) - ServerClient: Flush cached data (Michal Privoznik) - Fix storage pool source comparison to avoid comparing with self (Daniel P. Berrange) - qemu: plug memory leak (Alex Jia) - qemu: Restore the original states of PCI device when restarting daemon (Osier Yang) - macvtap: Fix error return value convention/inconsistencies (Roopa Prabhu) - pci address conflict when virtio disk with drive type (Xu He Jie) - qemu: plug memory leak (Alex Jia) - qemu: avoid leaking uninit data from hotplug to dumpxml (Eric Blake) - util: Fix virUUIDGeneratePseudoRandomBytes (Ryota Ozaki) - lxc: Revert zeroing count of allocated items if VIR_REALLOC_N fails (Peter Krempa) - lxc: avoid null deref on lxcSetupLoopDevices failure (Alex Jia) - lxc: avoid missing '{' in the function (Alex Jia) - storage: avoid null deref on qemu-img failure (Eric Blake) - storage: make previous leak less likely to regress (Eric Blake) - storage: plug iscsi memory leak (Eric Blake) - qemu: avoid leaking uninit data from hotplug to dumpxml (Eric Blake) - qemu: Do not wait if the PCI device is not managed when reattaching (Osier Yang) - Add missing strdup return value check (Roopa Prabhu) - macvtap: avoid invalid free (Roopa Prabhu) - util: Fix typo in virGetHostname description (Jiri Denemark) - macvtap: plug memory leak for 802.1Qbh (Eric Blake) - qemu: plug memory leak on migration (Eric Blake) - conf: plug memory leak on error (Eric Blake) - storage: plug memory leak on error (Eric Blake) - util: Make getaddrinfo failure nonfatal in virGetHostname (Jiri Denemark) - qemu: fix text block info parsing (Eric Blake) - qemu: avoid text monitor null deref (Eric Blake) - qemu: check for json allocation failure (Eric Blake) - virFDStream: close also given errfd (fd leak) (Marc-André Lureau) - command: avoid fd leak on failure (Eric Blake) - qemu: Check for domain being active on successful job acquire (Michal Privoznik) - xen: Return tap2 for tap2 disks (Philipp Hahn) - xen: fix PyGrub boot device order (Philipp Hahn) - build: fix 'make check' linkage with dtrace (Eric Blake) - Fix deps for probes.o to ensure correct build ordering (Daniel P. Berrange) - If receiving a stream error, mark EOF on the stream (Daniel P. Berrange) - xen_xs: Guard against set but empty kernel argument (Guido Günther) - snapshot: avoid accidental renames with snapshot-edit (Eric Blake) - storage: Do not use comma as seperator for lvs output (Osier Yang) - qemuDomainAttach: Initialize pidfile variable (Michal Privoznik) - lxc: fix logic bug (Eric Blake) - Don't send back unknown program errors for async messages (Daniel P. Berrange) - Fix deadlock when the RPC program is unknown (Daniel P. Berrange) - remote_driver: Avoid double free in EventControl building (Michal Privoznik) - xenParseXM: don't dereference NULL pointer when script is empty (Guido Günther) - qemu: Fix migration with dname (Jiri Denemark) - virsh: do not unlink NULL file (Marc-André Lureau) - qemu: Fix error message mentioning VNC instead of SPICE (Peter Krempa) - qemu: Check for ejected media during startup and migration (Michal Privoznik) - qemu: add return value check (Alex Jia) - qemu: Always remove domain object if MigratePrepare fails (Jiri Denemark) - fix AppArmor driver for pipe character devices (Jamie Strandboge) - daemon: Don't remove pidfiles in init scripts (Peter Krempa) - storage: Do not break the whole vol lookup process in the middle (Osier Yang) - Fix synchronous reading of stream data (Daniel P. Berrange) Improvements: - Add missing defaultConsoleTargetType callback for AppArmour (Daniel P. Berrange) - Fix naming of constant for disk event (Daniel P. Berrange) - lxc: use common code for process cleanup (Eric Blake) - Set aliases for LXC/UML console devices (Daniel P. Berrange) - Default console target type with no <target> element (Daniel P. Berrange) - Add support for multiple consoles in LXC (Daniel P. Berrange) - Rewrite LXC I/O forwarding to use main event loop (Daniel P. Berrange) - Allow multiple consoles per virtual guest (Daniel P. Berrange) - virnetsockettest: Use a temporary directory in /tmp (Guido Günther) - xen: allow getting < max typed parameters (Eric Blake) - lxc: allow getting < max typed parameters (Eric Blake) - libxl: allow getting < max typed parameters (Eric Blake) - esx: allow getting < max typed parameters (Eric Blake) - qemu: allow getting < max typed parameters (Eric Blake) - Add support for probing filesystem with libblkid (Daniel P. Berrange) - Fix error message when failing to detect filesystem (Daniel P. Berrange) - Workaround for broken kernel autofs mounts (Daniel P. Berrange) - Ensure errno is valid when returning from lxcContainerWaitForContinue (Daniel P. Berrange) - Create /var/lib/libvirt/filesystems for LXC trees (Daniel P. Berrange) - esx: Support folders in the path of vpx:// connection URIs (Matthias Bolte) - qemu: pass virConnectPtr into Domain{Attach,Detach}* (Sage Weil) - vbox: Support shared folders (Matthias Bolte) - xenapi: Improve error reporting in xenapiOpen once again (Matthias Bolte) - Use a common xml type for ceph secret usage. (Josh Durgin) - storage: add auth to virDomainDiskDef (Josh Durgin) - secret: add Ceph secret type (Sage Weil) - Implement RPC driver support for virDomainOpenGraphics (Daniel P. Berrange) - Extend RPC server to allow FD passing (Daniel P. Berrange) - Add client side support for FD passing (Daniel P. Berrange) - Extend RPC protocol to allow FD passing (Daniel P. Berrange) - Add APIs for virNetSocket for sending/receiving file descriptors (Daniel P. Berrange) - Wire up QEMU implementation for virDomainOpenGraphics (Daniel P. Berrange) - Extend graphics event to include UNIX socket (Daniel P. Berrange) - virsh: Fix error message on vol-create-from failure (Ryota Ozaki) - bridge: modify for use when sVirt is enabled with qemu (Tyler Coumbes) - Use virXMLSaveFile when writing XML config (Jiri Denemark) - Introduce virXMLSaveFile as a wrapper for virFileRewrite (Jiri Denemark) - Introduce virFileRewrite for safe file rewrite (Jiri Denemark) - Add a systemtap script for watching QEMU monitor interactions (Daniel P. Berrange) - qemu: simplify use of HAVE_YAJL (Eric Blake) - snapshot: simplify indentation of disk encryption xml (Eric Blake) - snapshot: simplify indentation of nwfilter (Eric Blake) - Add REMOTE_PROC_DOMAIN_EVENT_DISK_CHANGE to remote_protocol-structs (Daniel P. Berrange) - nwfilter: extend schema to support new targets (Stefan Berger) - util: Add virFileAccessibleAs to private symbols (Michal Privoznik) - startupPolicy: Emit event on disk source dropping (Michal Privoznik) - qemu: implement startupPolicy (Michal Privoznik) - qemu: Move device alias assigning before command line construction (Michal Privoznik) - util: Create virFileAccessibleAs function (Michal Privoznik) - conf: Introduce optional startupPolicy attribute for cdrom and floppy (Michal Privoznik) - waitpid: improve safety (Eric Blake) - virsh: Fix vol-info's 'Type' output (Ryota Ozaki) - support continue/return targets in nwfilter (David L Stevens) - snapshot: simplify indentation of network xml (Eric Blake) - snapshot: simplify indentation of cpu features (Eric Blake) - snapshot: simplify indentation of sysinfo (Eric Blake) - snapshot: test domainsnapshot indentation (Eric Blake) - snapshot: indent domain xml when nesting (Eric Blake) - virbuf: add auto-indentation support (Eric Blake) - virbuf: more detailed error reporting (Eric Blake) - virbuf: improve testsuite reporting (Eric Blake) - virbuf: fix const-correctness (Eric Blake) - qemu: allow json in domxml-to-native (tangchen) - support setting bandwidth from virsh attach-interface (Hu Tao) - lxc: use hand-rolled code in place of unlockpt and grantpt (Serge E. Hallyn) - qemu: Test name-space handling (Philipp Hahn) - qemu: Fix name-space handling (Philipp Hahn) - Replace virBufferAdd with virBufferAddLit for const string (Daniel P. Berrange) - Allow for URI aliases when connecting to libvirt (Daniel P. Berrange) - Add support for autodestroy of guests to the LXC and UML drivers (Daniel P. Berrange) - Use virBufferEscapeShell in cmdEcho (Guido Günther) - qemu: replace qemuMonitorEscapeShell by virBufferEscapeShell (Guido Günther) - virBufferEscapeShell: Emit quotes for the empty string (Guido Günther) - snapshot: detect when qemu lacks disk-snapshot support (Eric Blake) - virBufferEscapeShell: Fix escaping of single quotes. (Guido Günther) - compile: Add a missing function 'pciDeviceListFind' to libvirt_private.syms (Xu He Jie) - snapshot: implement LIST_LEAVES flag in esx (Eric Blake) - qemu: Relax -no-shutdown check to [0.14.0, 0.15.0] (Jiri Denemark) - qemu: Honor the orginal PCI dev properties when reattaching (Osier Yang) - qemu: Do not reattach PCI device used by other domain when shutdown (Osier Yang) - Xen: Fake versions in xencapstest (Philipp Hahn) - Xen: move versions to struct (Philipp Hahn) - Use virBufferEscapeShell in virNetSocketNewConnectSSH (Guido Günther) - Add virBufferEscapeShell (Guido Günther) - Autodetect if the remote nc command supports the -q option (Guido Günther) - qemu: Make sure BeginJob is always followed by EndJob (Jiri Denemark) - qemu: Log debug messages when changing job (Jiri Denemark) - build: add compiler attributes to virUUIDParse (Eric Blake) - Fix typo in lxc_controller (Serge E. Hallyn) - build: update to latest gnulib (Eric Blake) - events: Propose a separate lock for event queue (Michal Privoznik) - qemu: Implement VIR_DUMP_RESET (Michal Privoznik) - virDomainCoreDump: Introduce VIR_DUMP_RESET flag (Michal Privoznik) - example: Support debug output and loop switch (Philipp Hahn) - example: Redirect --help output to stdout/stderr (Philipp Hahn) - example: Fix argument handling (Philipp Hahn) - snapshot: implement LIST_LEAVES flag in qemu (Eric Blake) - snapshot: add API for filtering by leaves (Eric Blake) - tests: Add support for skipping tests (Philipp Hahn) - Introduce <driver> under <filesystem> to support open-by-handle (Harsh Prateek Bora) - buf: implement generic virBufferEscape (Sage Weil) - daemon: Always advertise libvirtd service (Osier Yang) - snapshot: take advantage of new relations (Eric Blake) - snapshot: track qemu snapshot relations (Eric Blake) - snapshot: framework for more efficient relation traversal (Eric Blake) - snapshot: use correct qmp monitor command (Eric Blake) - snapshot: virsh shorthand for operating on current snap (Eric Blake) - build: ship helper scripts (Eric Blake) - Rewrite all the DTrace/SystemTAP probing (Daniel P. Berrange) - Fix missing lock calls on virNetTLSContextRef (Daniel P. Berrange) - Refactor TLS to facilitate dynamic probing (Daniel P. Berrange) - Add virSocketRef API to facilitate dynamic probing (Daniel P. Berrange) - Make libvirt.so include the RPC server code (Daniel P. Berrange) - snapshot: implement snapshot children listing in esx (Eric Blake) - snapshot: implement snapshot children listing in qemu (Eric Blake) - snapshot: remote protocol for snapshot children (Eric Blake) - snapshot: virsh fallback for snapshot-list --descendants --from (Eric Blake) - snapshot: virsh fallback for snapshot-list --from children (Eric Blake) - snapshot: virsh fallback for snapshot-list --tree --from (Eric Blake) - snapshot: virsh snapshot-list and children (Eric Blake) - xen: add error handling to UUID parsing (Guido Günther) - maint: typo fixes (Eric Blake) - snapshot: sort snapshot-list --tree (Eric Blake) - snapshot: simplify redefinition of disk snapshot (Eric Blake) - snapshot: let virsh edit disk snapshots (Eric Blake) - snapshot: fix virsh error message typo (Eric Blake) - qemu: add separate rerror_policy for disk errors (Laine Stump) - qemu: leave rerror policy at default when enospace is requested (Laine Stump) - qemu: enable multifunction for older qemu (Eric Blake) - Make LXC work with new network configuration types (Daniel P. Berrange) - init: raise default system aio limits (Eric Blake) - maint: fix minor issues in virterror public header (Eric Blake) - snapshot: enforce REVERT_FORCE on qemu (Eric Blake) - snapshot: use qemu-img on disks in use at time of snapshot (Eric Blake) - snapshot: add REVERT_FORCE to API (Eric Blake) - snapshot: implement snapshot roots listing in vbox (Eric Blake) - qemu: Don't fail virDomainGetInfo if we can't update balloon info (Jiri Denemark) - snapshot: simplify esx snapshot name lookup (Eric Blake) - snapshot: implement snapshot roots listing in esx (Eric Blake) - qemu: correct misspelled 'enospc' option, and only use for werror (Laine Stump) - snapshot: better virsh handling of missing current, parent (Eric Blake) - Allow passing of command line args to LXC container (Daniel P. Berrange) - Add support for bandwidth filtering on LXC guests (Daniel P. Berrange) - network: fill in bandwidth from portgroup for all forward modes (Laine Stump) - bridge_driver.c: Fix autoconf setting (Neil Wilson) - Set to NULL members that have been freed to prevent crashes (Marc-André Lureau) - snapshot: implement getparent for vbox (Eric Blake) - snapshot: implement getparent for esx (Eric Blake) - qemu: make PCI multifunction support more manual (Laine Stump) - lvm storage backend: handle command_names=1 in lvm.conf (Serge E. Hallyn) - qemu: Check for outstanding async job too (Michal Privoznik) - virsh: Add 'reset' command for virsh (Xu He Jie) - remote: Implement 'reset' for remote driver (Xu He Jie) - qemu: Implement 'reset' for qemu driver (Xu He Jie) - logging: Add date to log timestamp (Jiri Denemark) - logging: Do not log timestamp through syslog (Jiri Denemark) - hyperv: Report an error for acceptable URI schemes with a transport (Matthias Bolte) - esx: Report an error for acceptable URI schemes with a transport (Matthias Bolte) - snapshot: implement getparent in qemu (Eric Blake) - snapshot: add virsh snapshot-list --tree (Eric Blake) - snapshot: refactor virsh snapshot parent computation (Eric Blake) - snapshot: remote protocol for getparent (Eric Blake) - security: properly chown/label bidirectional and unidirectional fifos (Laine Stump) - qemu: Preserve fakeReboot flag in domain status (Jiri Denemark) - qemu: Finish domain shutdown on reconnect (Jiri Denemark) - qemu: Check domain status details when reconnecting monitor (Jiri Denemark) - virsh: Allow using complete <capabilities> elements with cpu-baseline (Peter Krempa) - virsh: Allow using domain and capabilities XMLs with cpu-compare (Peter Krempa) - qemu: add ability to set PCI device "rombar" on or off (Laine Stump) - qemu: Relax -no-shutdown check to [0.14.0, 0.15.50) (Jiri Denemark) - virLockManagerNopInit: Rename flags to unused_flags (Michal Privoznik) - debug: Annotate some variables as unused (Michal Privoznik) - maint: update authors (Peter Krempa) - Add unsafe cache mode support for disk driver (Oskari Saarenmaa) - selinux: Correctly report warning if virt_use_nfs not set (Michal Privoznik) - virsh: Do not ignore the specified flags for cmdSaveImageDefine (Osier Yang) - qemu: Transfer inactive XML among cookie (Michal Privoznik) Cleanups: - Remove translations in socket test case (Daniel P. Berrange) - test: drop redundant check (Eric Blake) - startupPolicty: Minor cleanups (Michal Privoznik) - secret: fix bad patch application (Eric Blake) - Remove trailing whitespace from all xfig files (Daniel P. Berrange) - Fix typo in virFileAccessibleAs (Daniel P. Berrange) - snapshot: minor cleanups from reviewing indentation (Eric Blake) - build: ignore test executable (Eric Blake) - esx: drop dead code to silence Coverity (Eric Blake) - snapshot: drop dead parameters (Eric Blake) - qemu: silence Coverity false positive (Eric Blake) - conf: remove unused VIR_ENUM_DECL (Laine Stump) thanks to everybody who helped for this release, by reporting bugs, providing ideas, patches, or documentation ! 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/

1 0

[libvirt] [PATCH] [configure] Fix libvirt dependency, and use $VAPIGEN
by nirbheek＠gentoo.org 08 Nov '11

08 Nov '11

This is a rather trivial patch to raise the minimum libvirt dependency; at least 0.9.4 is required. In addition, it fixes the direct usage of `vapigen`, and uses the result of the AC_PATH_PROG(VAPIGEN ...) call earlier so that the binary to be used can be overridden as that macro was meant to allow. Thanks! configure.ac | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)

5 6

[libvirt] [PATCH] Add missing defaultConsoleTargetType callback for AppArmour
by Daniel P. Berrange 07 Nov '11

07 Nov '11

From: "Daniel P. Berrange" <berrange(a)redhat.com> Every instance of virCapsPtr must have the defaultConsoleTargetType field set. * src/security/virt-aa-helper.c: Add defaultConsoleTargetType to virCapsPtr --- src/security/virt-aa-helper.c | 7 +++++++ 1 files changed, 7 insertions(+), 0 deletions(-) diff --git a/src/security/virt-aa-helper.c b/src/security/virt-aa-helper.c index 71a4586..e8e77ba 100644 --- a/src/security/virt-aa-helper.c +++ b/src/security/virt-aa-helper.c @@ -685,6 +685,11 @@ caps_mockup(vahControl * ctl, const char *xmlStr) return rc; } +static int aaDefaultConsoleType(const char *ostype ATTRIBUTE_UNUSED) +{ + return VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL; +} + static int get_definition(vahControl * ctl, const char *xmlStr) { @@ -703,6 +708,8 @@ get_definition(vahControl * ctl, const char *xmlStr) goto exit; } + caps->defaultConsoleTargetType = aaDefaultConsoleType; + if ((guest = virCapabilitiesAddGuest(ctl->caps, ctl->hvm, ctl->arch, -- 1.7.6.4

2 1

[libvirt] Failing pool deletion testcase
by Guido Günther 07 Nov '11

07 Nov '11

Hi, when running libvirt-tck I'm seeing this test failure: # Failed test 'deleted pool' # at /usr/share/libvirt-tck/tests/storage/110-disk-pool.t line 120. # died: Sys::Virt::Error (libvirt error code: 3, message: this function is not supported by the connection driver: pool does not support pool deletion # ) because pool deletion isn't supported by the disk storage backend. Since the test is there I wonder if there are patches floating around already to implement it? If not I guess wiping the partition table followed by a "blockdev --reread-pt" would be the right approach? Cheers, -- Guido

2 2

[libvirt] [PATCH] libvirt-tck: Skip Test when no security model is in use
by Guido Günther 07 Nov '11

07 Nov '11

Hi, if I'm reading this test correctly it's supposed to fail without a security model at work. The file images are always readable/writeable to the libvirt user and group without one. Cheers, -- Guido --- scripts/qemu/205-qcow2-double-backing-file.t | 9 +++++++-- 1 files changed, 7 insertions(+), 2 deletions(-) diff --git a/scripts/qemu/205-qcow2-double-backing-file.t b/scripts/qemu/205-qcow2-double-backing-file.t index 7ae278d..d3a5e33 100644 --- a/scripts/qemu/205-qcow2-double-backing-file.t +++ b/scripts/qemu/205-qcow2-double-backing-file.t @@ -52,17 +52,22 @@ my $conn = eval { $tck->setup(); }; BAIL_OUT "failed to setup test harness: $@" if $@; END { $tck->cleanup if $tck; } +my $info; +eval { + $info = $conn->get_node_security_model(); +}; + SKIP: { skip "Only relevant to QEMU driver", 26 unless $conn->get_type() eq "QEMU"; skip "Only relevant when run as root", 26 unless $< == 0; skip "Only relevant for system driver", 26 unless $conn->get_uri() =~ m/system/; - + skip "Only relevant when using a security model", 26 unless + $info && $info->{model}; my $xml = $tck->generic_pool("dir") ->mode("0755")->as_xml; - diag "Defining transient storage pool $xml"; my $pool; ok_pool(sub { $pool = $conn->define_storage_pool($xml) }, "define transient storage pool"); -- 1.7.7

2 2

[libvirt] [PATCH] Add ability to managed save libvirt domains
by Pavel Odintsov 07 Nov '11

07 Nov '11

--- src/libvirt-php.c | 21 +++++++++++++++++++++ src/libvirt-php.h | 1 + 2 files changed, 22 insertions(+), 0 deletions(-) diff --git a/src/libvirt-php.c b/src/libvirt-php.c index f590595..1c39e31 100644 --- a/src/libvirt-php.c +++ b/src/libvirt-php.c @@ -91,6 +91,7 @@ static function_entry libvirt_functions[] = { PHP_FE(libvirt_domain_core_dump, NULL) PHP_FE(libvirt_domain_shutdown, NULL) PHP_FE(libvirt_domain_suspend, NULL) + PHP_FE(libvirt_domain_managedsave, NULL) PHP_FE(libvirt_domain_undefine, NULL) PHP_FE(libvirt_domain_reboot, NULL) PHP_FE(libvirt_domain_define_xml, NULL) @@ -3625,6 +3626,26 @@ PHP_FUNCTION(libvirt_domain_suspend) } /* + Function name: libvirt_domain_managedsave + Since version: 0.4.1(-1) + Description: Function is used to managed save the domain (domain was unloaded from memory and its state was saved to disk) identified by it's resource + Arguments: @res [resource]: libvirt domain resource, e.g. from libvirt_domain_lookup_by_*() + Returns: TRUE for success, FALSE on error +*/ +PHP_FUNCTION(libvirt_domain_managedsave) +{ + php_libvirt_domain *domain=NULL; + zval *zdomain; + int retval; + + GET_DOMAIN_FROM_ARGS("r",&zdomain); + retval=virDomainManagedSave(domain->domain, 0); + DPRINTF("%s: virDomainManagedSave(%p) returned %d\n", PHPFUNC, domain->domain, retval); + if (retval != 0) RETURN_FALSE; + RETURN_TRUE; +} + +/* Function name: libvirt_domain_undefine Since version: 0.4.1(-1) Description: Function is used to undefine the domain identified by it's resource diff --git a/src/libvirt-php.h b/src/libvirt-php.h index 2addae2..8a3f4ff 100644 --- a/src/libvirt-php.h +++ b/src/libvirt-php.h @@ -237,6 +237,7 @@ PHP_FUNCTION(libvirt_domain_resume); PHP_FUNCTION(libvirt_domain_core_dump); PHP_FUNCTION(libvirt_domain_shutdown); PHP_FUNCTION(libvirt_domain_suspend); +PHP_FUNCTION(libvirt_domain_managedsave); PHP_FUNCTION(libvirt_domain_undefine); PHP_FUNCTION(libvirt_domain_reboot); PHP_FUNCTION(libvirt_domain_define_xml); -- 1.7.2.5

1 0

[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