- Devel - libvirt List Archives

[libvirt] [PATCH] check return values of virBufferTrim
by Ján Tomko 18 Jun '13

18 Jun '13

Just to silence Coverity: Event check_return: Calling function "virBufferTrim(virBufferPtr, char const *, int)" without checking return value (as is done elsewhere 5 out of 6 times). --- src/node_device/node_device_udev.c | 5 ++--- src/rpc/virnetsshsession.c | 3 +-- 2 files changed, 3 insertions(+), 5 deletions(-) diff --git a/src/node_device/node_device_udev.c b/src/node_device/node_device_udev.c index bb58415..a37989a 100644 --- a/src/node_device/node_device_udev.c +++ b/src/node_device/node_device_udev.c @@ -370,9 +370,8 @@ udevLogFunction(struct udev *udev ATTRIBUTE_UNUSED, const char *format = NULL; virBufferAdd(&buf, fmt, -1); - virBufferTrim(&buf, "\n", -1); - - format = virBufferContentAndReset(&buf); + if (virBufferTrim(&buf, "\n", -1) >= 0) + format = virBufferContentAndReset(&buf); virLogVMessage(VIR_LOG_FROM_LIBRARY, virLogPriorityFromSyslog(priority), diff --git a/src/rpc/virnetsshsession.c b/src/rpc/virnetsshsession.c index b6aedc8..2299871 100644 --- a/src/rpc/virnetsshsession.c +++ b/src/rpc/virnetsshsession.c @@ -362,9 +362,8 @@ virNetSSHCheckHostKey(virNetSSHSessionPtr sess) * we have to use a *MAGIC* constant. */ for (i = 0; i < 16; i++) virBufferAsprintf(&buff, "%02hhX:", keyhash[i]); - virBufferTrim(&buff, ":", 1); - if (virBufferError(&buff) != 0) { + if (virBufferTrim(&buff, ":", 1) < 0) { virReportOOMError(); return -1; } -- 1.8.1.5

3 4

[libvirt] Not able to open Xen Connection using libvirt
by varun bhatnagar 18 Jun '13

18 Jun '13

Hi, I am trying to create virtual machine through libvirt using xen hypervisor. I was able to install xen successfully but when I tried connecting to the hypervisor it gave an error. And while installing libvirt (libvirt-1.0.1) I installed it with xen support. I am pasting the snapshot of the error below: [image: Inline image 1] Can anyone tell me why I am getting this error? // Varun

1 0

[libvirt] [PATCH 0/2] Don't always fail migrations on I/O errors
by Peter Krempa 18 Jun '13

18 Jun '13

Recently I've added code that aborts migration in case of I/O error. This may not be desirable as qemu does actually support such migration. This series adds a flag that enables this option that will be now disabled by default. Peter Krempa (2): migration: Make erroring out on I/O error controllable by flag migration: Don't propagate VIR_MIGRATE_ABORT_ON_ERROR include/libvirt/libvirt.h.in | 1 + src/libvirt.c | 17 +++++++++++------ src/qemu/qemu_driver.c | 4 ++-- src/qemu/qemu_migration.c | 39 ++++++++++++++++++++++++--------------- src/qemu/qemu_migration.h | 6 ++++-- tools/virsh-domain.c | 7 +++++++ tools/virsh.pod | 6 ++++-- 7 files changed, 53 insertions(+), 27 deletions(-) -- 1.8.2.1

3 4

[libvirt] [PATCH] qemu: Avoid leaking uri in qemuMigrationPrepareDirect
by Jiri Denemark 18 Jun '13

18 Jun '13

--- src/qemu/qemu_migration.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index 48e0d44..68c9cc5 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -2367,7 +2367,7 @@ qemuMigrationPrepareDirect(virQEMUDriverPtr driver, const char *p; char *uri_str = NULL; int ret = -1; - virURIPtr uri; + virURIPtr uri = NULL; VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, " "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, " @@ -2472,6 +2472,7 @@ qemuMigrationPrepareDirect(virQEMUDriverPtr driver, cookieout, cookieoutlen, def, NULL, this_port, flags); cleanup: + virURIFree(uri); VIR_FREE(hostname); if (ret != 0) VIR_FREE(*uri_out); -- 1.8.2.1

2 2

[libvirt] [PATCH 1/2 v3] json: add stream parser
by Dmitry Guryanov 18 Jun '13

18 Jun '13

Add function virJSONValueFromStream, which reads data from a stream and passes it to json parser. When end of the object is reached, it returns this object. To avoid reading from the stream by single bytes it reads to a buffer (in a structure virJSONStreamParserState), which should be passed to a consequent call of this function. So if the end of one object and the beginning of the next object have been read by a single system call - virJSONValueFromStream handle it correctly. example of usage: virJSONValuePtr v; virJSONStreamParserState state; memset(&state, 0, sizeof(state)); while (1) { v = virJSONValueFromStream(mon->fd, &state); if (v == (void *)-1) /* error */ break; if (v == NULL) /* file descriptor has been closed */ break; /* handle object 'v' */ } I need such function for the parallels driver. It caches info about domains and needs some mechanism to update this cache. There is a "prlsrvctl monitor" command which waits for events forever and prints info about events to stdout in json format. So parallels driver could start separate thread which will read from prlsrvctl's stdout and update cache accordingly. There is the same task in qemu_monitor_json, but each json object is printed in a separate line there. It's not possible in my case, because some fields could have line endings. --- src/libvirt_private.syms | 1 + src/util/virjson.c | 110 ++++++++++++++++++++++++++++++++++++++++++++++ src/util/virjson.h | 8 +++ 3 files changed, 119 insertions(+), 0 deletions(-) * This patch left unchanged since v2 diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms index b449293..0db2a0d 100644 --- a/src/libvirt_private.syms +++ b/src/libvirt_private.syms @@ -1391,6 +1391,7 @@ virJSONValueArrayGet; virJSONValueArraySize; virJSONValueFree; virJSONValueFromString; +virJSONValueFromStream; virJSONValueGetBoolean; virJSONValueGetNumberDouble; virJSONValueGetNumberInt; diff --git a/src/util/virjson.c b/src/util/virjson.c index 3a6f520..9559a26 100644 --- a/src/util/virjson.c +++ b/src/util/virjson.c @@ -1022,6 +1022,116 @@ cleanup: return ret; } +/* + * Read single JSON object from the stream. Store data, which + * have already been read from the stream, but belongs to the + * next object to the virJSONStreamParserState structure. So that + * consequent call to this function will parse it and return that + * object. + * + * @fd: file descriptor, opened for reading + * @state: pointer to the structure with buffer for data, read from fd. + * + * Return (void *)-1 in case of error, NULL when eof reached, + * pointer to the virJSONValuePtr in case of success. + */ + +virJSONValuePtr virJSONValueFromStream(int fd, virJSONStreamParserStatePtr state) +{ + yajl_handle hand; + virJSONParser parser = { NULL, NULL, 0 }; + virJSONValuePtr value = (void *)-1; +# ifndef WITH_YAJL2 + yajl_parser_config cfg = { 1, 1 }; +# endif + ssize_t len; + int ret = 0; + bool done = false; + +# ifdef WITH_YAJL2 + hand = yajl_alloc(&parserCallbacks, NULL, &parser); + if (hand) { + yajl_config(hand, yajl_allow_comments, 1); + yajl_config(hand, yajl_dont_validate_strings, 0); + yajl_config(hand, yajl_allow_trailing_garbage, 1); + } +# else + hand = yajl_alloc(&parserCallbacks, &cfg, NULL, &parser); +# endif + if (!hand) { + virReportError(VIR_ERR_INTERNAL_ERROR, "%s", + _("Unable to create JSON parser")); + goto cleanup; + } + + do { + if (strlen(&state->buf[state->pos]) == 0) { + state->pos = 0; + memset(state->buf, 0, sizeof(state->buf)); + len = read(fd, state->buf, sizeof(state->buf) - 1); + + if (len < 0) { + virReportSystemError(errno, _("cannot read from fd '%d'"), fd); + virJSONValueFree(parser.head); + goto cleanup; + } + + if (len == 0) { + value = parser.head; + goto cleanup; + } + } + + for (;state->pos < strlen(state->buf); state->pos++) { + unsigned char *buf = (unsigned char *)&state->buf[state->pos]; + /* + * New yaml library has useful function yajl_get_bytes_consumed + * which allows parsing by larger chunks. But rhel-6 has 1.0.7 + * version, which doesn't have it. + */ + ret = yajl_parse(hand, buf, 1); +# ifdef WITH_YAJL2 + if (ret == 0 && yajl_get_bytes_consumed(hand) == 0) { + done = true; + /* state->pos points to the first symbol after current + * object */ + break; + } else if (ret != 0) { +# else + if (ret == 0) { + done = true; + /* state->pos points to the last symbol of the + * current object */ + state->pos++; + break; + } else if (ret != yajl_status_insufficient_data) { +# endif + unsigned char *errstr = yajl_get_error(hand, 1, buf, 1); + virReportError(VIR_ERR_INTERNAL_ERROR, + _("cannot parse json: %s"), + (const char*) errstr); + VIR_FREE(errstr); + virJSONValueFree(parser.head); + goto cleanup; + } + } + } while (!done); + + value = parser.head; + +cleanup: + yajl_free(hand); + + if (parser.nstate) { + int i; + for (i = 0 ; i < parser.nstate ; i++) { + VIR_FREE(parser.state[i].key); + } + } + + VIR_DEBUG("result=%p", parser.head); + return value; +} static int virJSONValueToStringOne(virJSONValuePtr object, yajl_gen g) diff --git a/src/util/virjson.h b/src/util/virjson.h index db11396..1f5ce38 100644 --- a/src/util/virjson.h +++ b/src/util/virjson.h @@ -48,6 +48,8 @@ typedef virJSONObjectPair *virJSONObjectPairPtr; typedef struct _virJSONArray virJSONArray; typedef virJSONArray *virJSONArrayPtr; +typedef struct _virJSONStreamParserState virJSONStreamParserState; +typedef virJSONStreamParserState *virJSONStreamParserStatePtr; struct _virJSONObjectPair { char *key; @@ -77,6 +79,11 @@ struct _virJSONValue { } data; }; +struct _virJSONStreamParserState { + char buf[1024]; + size_t pos; +}; + void virJSONValueFree(virJSONValuePtr value); virJSONValuePtr virJSONValueNewString(const char *data); @@ -138,5 +145,6 @@ int virJSONValueObjectRemoveKey(virJSONValuePtr object, const char *key, virJSONValuePtr virJSONValueFromString(const char *jsonstring); char *virJSONValueToString(virJSONValuePtr object, bool pretty); +virJSONValuePtr virJSONValueFromStream(int fd, virJSONStreamParserStatePtr state); #endif /* __VIR_JSON_H_ */ -- 1.7.1

1 1

[libvirt] [PATCH] qemu_migration: Move waiting for SPICE migration
by Michal Privoznik 18 Jun '13

18 Jun '13

Currently, we wait for SPICE to migrate in the very same loop where we wait for qemu to migrate. This has a disadvantage of slowing seamless migration down. One one hand, we should not kill the domain until all SPICE data has been migrated. On the other hand, there is no need to wait in the very same loop and hence slowing down 'cont' on the destination. For instance, if users are watching a movie, they can experience the movie to be stopped for a couple of seconds, as processors are not running nor on src nor on dst as libvirt waits for SPICE to migrate. We should move the waiting phase to migration CONFIRM phase. --- src/qemu/qemu_migration.c | 57 ++++++++++++++++++++++++++++++++--------------- 1 file changed, 39 insertions(+), 18 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index ca79bc2..1d65df5 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -1582,6 +1582,40 @@ cleanup: return ret; } +static int +qemuMigrationWaitForSpice(virQEMUDriverPtr driver, + virDomainObjPtr vm) +{ + qemuDomainObjPrivatePtr priv = vm->privateData; + bool wait_for_spice = false; + bool spice_migrated = false; + + if (vm->def->ngraphics == 1 && + vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE && + virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION)) + wait_for_spice = true; + + if (!wait_for_spice) + return 0; + + while (!spice_migrated) { + /* Poll every 50ms for progress & to allow cancellation */ + struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull }; + + qemuDomainObjEnterMonitor(driver, vm); + if (qemuMonitorGetSpiceMigrationStatus(priv->mon, + &spice_migrated) < 0) { + qemuDomainObjExitMonitor(driver, vm); + return -1; + } + qemuDomainObjExitMonitor(driver, vm); + virObjectUnlock(vm); + nanosleep(&ts, NULL); + virObjectLock(vm); + } + + return 0; +} static int qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver, @@ -1591,19 +1625,10 @@ qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver, { qemuDomainObjPrivatePtr priv = vm->privateData; int ret; - bool wait_for_spice = false; - bool spice_migrated = false; qemuMonitorMigrationStatus status; memset(&status, 0, sizeof(status)); - /* If guest uses SPICE and supports seamles_migration we have to hold up - * migration finish until SPICE server transfers its data */ - if (vm->def->ngraphics == 1 && - vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE && - virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION)) - wait_for_spice = true; - ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob); if (ret < 0) { /* Guest already exited; nothing further to update. */ @@ -1611,13 +1636,6 @@ qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver, } ret = qemuMonitorGetMigrationStatus(priv->mon, &status); - /* If qemu says migrated, check spice */ - if (wait_for_spice && - ret == 0 && - status.status == QEMU_MONITOR_MIGRATION_STATUS_COMPLETED) - ret = qemuMonitorGetSpiceMigrationStatus(priv->mon, - &spice_migrated); - qemuDomainObjExitMonitor(driver, vm); priv->job.status = status; @@ -1657,8 +1675,7 @@ qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver, break; case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED: - if ((wait_for_spice && spice_migrated) || (!wait_for_spice)) - priv->job.info.type = VIR_DOMAIN_JOB_COMPLETED; + priv->job.info.type = VIR_DOMAIN_JOB_COMPLETED; ret = 0; break; @@ -4121,6 +4138,10 @@ int qemuMigrationConfirm(virQEMUDriverPtr driver, * domain object, but if no, resume CPUs */ if (retcode == 0) { + /* If guest uses SPICE and supports seamless migration we have to hold + * up domain shutdown until SPICE server transfers its data */ + qemuMigrationWaitForSpice(driver, vm); + qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED, VIR_QEMU_PROCESS_STOP_MIGRATED); virDomainAuditStop(vm, "migrated"); -- 1.8.1.5

2 2

[libvirt] [PATCH 0/6] Enumerate scsi generic device
by Osier Yang 18 Jun '13

18 Jun '13

The scsi generic device is listed as a child of the parent of the mapped disk device. E.g. +- pci_0000_00_1f_2 | | | +- scsi_host0 | | | | | +- scsi_target0_0_0 | | | | | +- scsi_0_0_0_0 | | | | | +- block_sda_TOSHIBA_MK5061GSY_9243PG8CT | | +- scsi_generic_sg0 This is consistent with the sysfs/udev. The XML produced by udev backend: <device> <name>scsi_generic_sg0</name> <path>/sys/devices/pci0000:00/0000:00:1f.2/ata1/host0/target0:0:0/0:0:0:0/scsi_generic/sg0</path> <parent>scsi_0_0_0_0</parent> <capability type='scsi_generic'> <char>/dev/sg0</char> </capability> </device> The XML produced by HAL backend: <device> <name>pci_8086_2922_scsi_host_scsi_device_lun0_scsi_generic</name> <path>/sys/devices/pci0000:00/0000:00:1f.2/host0/target0:0:0/0:0:0:0/scsi_generic/sg0</path> <parent>pci_8086_2922_scsi_host_scsi_device_lun0</parent> <capability type='scsi_generic'> <char>/dev/sg0</char> </capability> </device> Osier Yang (6): nodedev: Expose sysfs path of device nodedev_udev: Refactor udevGetDeviceType nodedev_udev: Enumerate scsi generic device nodedev_hal: Enumerate scsi generic device nodedev: Support SCSI_GENERIC cap flag for listAllNodeDevices virsh: Support SCSI_GENERIC cap flag for nodedev-list include/libvirt/libvirt.h.in | 1 + src/conf/node_device_conf.c | 11 ++- src/conf/node_device_conf.h | 8 ++- src/node_device/node_device_hal.c | 9 +++ src/node_device/node_device_udev.c | 137 ++++++++++++++++++++----------------- tools/virsh-nodedev.c | 3 + tools/virsh.pod | 6 +- 7 files changed, 105 insertions(+), 70 deletions(-) -- 1.8.1.4

2 17

[libvirt] a problem about the progress's display of migration with non-shared storage
by Liuji (Jeremy) 18 Jun '13

18 Jun '13

Hi all, I found a problem about the progress's display of migration with non-shared storage. In the commit: 7b7600b3e6734117b3db39ba5c31ae33a40fb5bb (qemu_migration: Introduce qemuMigrationDriveMirror), it use drive mirror instead of block migration. And after the driver mirror, then process the VM migration(no block). But only in the VM migration stage, libvirt set the progress of migration(in qemuMigrationWaitForCompletion). In the drive mirror stage, libvirt is waiting the completion of drive mirror, and don't set the progress of migration. So, the libvirt can't display the progress of migration in the drive mirror stage. For users, they are not concerned about the implementation details of the two stages. They think the two stages are both migration. So, I think libvirt need consider the progress of drive mirror for migration. But I don't have a good solution. Any idea about this problem? Thanks for advance Regards

2 1

[libvirt] [PATCH 0/6] Support to use iscsi storage in domain conf
by Osier Yang 18 Jun '13

18 Jun '13

This supports to use libvirt iscsi storage volume for qemu with either the LUN's path on host (e.g. /dev/disk/by-path/*-lun-1), or the the libiscsi uri (e.g iscsi://demo.org:6000/$iqn/1)in domain conf. Osier Yang (6): storage_iscsi: Reflect the default target port conf: Introduce new XML tag "mode" for disk source qemu: Translate the iscsi pool/volume disk source security: Ignore to manage the volume type disk if its mode is uri conf: Ignore the volume type disk if its mode is "uri" qemu: Translate the volume type disk source before cgroup setting docs/formatdomain.html.in | 9 ++- docs/schemas/domaincommon.rng | 8 +++ src/conf/domain_conf.c | 71 +++++++++++++++++++++- src/conf/domain_conf.h | 25 ++++++++ src/libvirt_private.syms | 1 + src/qemu/qemu_command.c | 20 ++++-- src/qemu/qemu_conf.c | 117 +++++++++++++++++++++++++++++------- src/qemu/qemu_process.c | 13 ++-- src/security/security_apparmor.c | 10 ++- src/security/security_dac.c | 10 ++- src/security/security_selinux.c | 10 ++- src/storage/storage_backend_iscsi.c | 6 +- tests/qemuxml2xmltest.c | 1 + 13 files changed, 261 insertions(+), 40 deletions(-) -- 1.8.1.4

1 6

[libvirt] [PATCH] Move virGetUserEnt() to where its needed
by Doug Goldstein 17 Jun '13

17 Jun '13

In the first if case, virGetUserEnt() isn't necessary so don't bother calling it before determining we need it. --- Found this trying to get libvirtd running happily on my Mac OS X machine for qemu. Unfortunately it appears virGetUserEnt() is always failing on Mac OS X (getpwuid_r() returns 0 each time) because we are requesting info on a different high value UIDs each time (32xxx). That's another issue entirely but this fix allows me to ignore that and test other fixes on my Mac. --- src/util/virutil.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/util/virutil.c b/src/util/virutil.c index c5246bc..6fa0212e 100644 --- a/src/util/virutil.c +++ b/src/util/virutil.c @@ -759,12 +759,13 @@ static char *virGetXDGDirectory(const char *xdgenvname, const char *xdgdefdir) { const char *path = getenv(xdgenvname); char *ret = NULL; - char *home = virGetUserEnt(geteuid(), VIR_USER_ENT_DIRECTORY); + char *home = NULL; if (path && path[0]) { if (virAsprintf(&ret, "%s/libvirt", path) < 0) goto no_memory; } else { + home = virGetUserEnt(geteuid(), VIR_USER_ENT_DIRECTORY); if (virAsprintf(&ret, "%s/%s/libvirt", home, xdgdefdir) < 0) goto no_memory; } -- 1.8.1.5

2 1