- Devel - libvirt List Archives

[libvirt] [PATCH] Fix disability to run on systems with no PCI bus
by Michal Privoznik 22 Apr '11

22 Apr '11

The patch which moved libpciaccess initialization to one place caused regression - we were not able to run on system with no PCI bus, like s390(x). This patch fix it. --- src/node_device/node_device_udev.c | 6 ++++++ 1 files changed, 6 insertions(+), 0 deletions(-) diff --git a/src/node_device/node_device_udev.c b/src/node_device/node_device_udev.c index 2139ef3..b6335ff 100644 --- a/src/node_device/node_device_udev.c +++ b/src/node_device/node_device_udev.c @@ -1603,7 +1603,13 @@ static int udevDeviceMonitorStartup(int privileged) char ebuf[256]; VIR_ERROR(_("Failed to initialize libpciaccess: %s"), virStrerror(pciret, ebuf, sizeof ebuf)); +#if defined __s390__ || defined __s390x_ + /* On x390(x) system there is no PCI bus. + * Therefore we do not want to return error here */ + ret = 0; +#else ret = -1; +#endif goto out; } } -- 1.7.4.4

4 3

[libvirt] [PATCH] Add support for s390(x) cpu options parsing
by Michal Privoznik 22 Apr '11

22 Apr '11

Up to now we missed parser for cpuinfo on x390(x) machines. Those machines have only 1 thread, core, socket. What is missing is information about CPU frequency. --- src/nodeinfo.c | 29 +++++++++++++++++++++++++++-- 1 files changed, 27 insertions(+), 2 deletions(-) diff --git a/src/nodeinfo.c b/src/nodeinfo.c index e0221f0..f55c83e 100644 --- a/src/nodeinfo.c +++ b/src/nodeinfo.c @@ -165,8 +165,10 @@ static int parse_socket(unsigned int cpu) { int ret = get_cpu_value(cpu, "topology/physical_package_id", false); # if defined(__powerpc__) || \ - defined(__powerpc64__) - /* ppc has -1 */ + defined(__powerpc64__) || \ + defined(__s390__) || \ + defined(__s390x__) + /* ppc and s390(x) has -1 */ if (ret < 0) ret = 0; # endif @@ -265,6 +267,29 @@ int linuxNodeInfoCPUPopulate(FILE *cpuinfo, && (*p == '\0' || *p == '.' || c_isspace(*p))) nodeinfo->mhz = ui; } +# elif defined(__s390__) || \ + defined(__s390x__) + } else if (STRPREFIX(buf, "# processors")) { + char *p; + unsigned int ui; + buf += 12; + while (*buf && c_isspace(*buf)) + buf++; + if (*buf != ':' || !buf[1]) { + nodeReportError(VIR_ERR_INTERNAL_ERROR, + _("parsing number of processors %c"), *buf); + return -1; + } + if (virStrToLong_ui(buf+1, &p, 10, &ui) == 0 + && (*p == '\0' || c_isspace(*p))) + nodeinfo->cpus = ui; + /* No other interesting infos are available in /proc/cpuinfo. + * However, there is a line identifying processor's version, + * identification and machine, but we don't want it to be caught + * and parsed in next iteration, because it is not in expected + * format and thus lead to error. */ + break; + } # else # warning Parser for /proc/cpuinfo needs to be adapted for your architecture # endif -- 1.7.4.4

3 2

Re: [libvirt] [PATCHv11 2/6] libvirt/qemu - clean up At(De)tachDeviceFlags() for consolidation.
by Eric Blake 22 Apr '11

22 Apr '11

On 04/21/2011 01:23 AM, KAMEZAWA Hiroyuki wrote: > clean up At(De)tachDeviceFlags() for consolidation. > > qemudDomainAttachDeviceFlags()/qemudDomainDetachFlags()/ > qemudDomainUpdateDeviceFlags() has similar logics and copied codes. > > This patch series tries to unify them to use shared code when it can. > At first, clean up At(De)tachDeviceFlags() and devide it into functions. > > By this, this patch pulls out shared components between functions. > Based on patch series by Eric Blake, I added some modification as > switch-case with QEMUD_DEVICE_ATTACH, QEMUD_DEVICE_DETACH, QEMUD_DEVICE_UPDATE I kind of liked my callback function pointers, but your switch statement isn't too many more lines of code and arguably a tiny bit more readable. I want this series in before 0.9.1, so I'm not going to reject the patch just because of a difference in approach :) > > +static int qemudDomainAttachDeviceDiskLive(struct qemud_driver *driver, The prefix qemud is a misnomer (it dates back to when we were running a daemon just for qemu and directly calling into xen; but now that the code has evolved, the daemon is named libvirtd and is independent of qemu, and the qemu code is independent of the daemon and no longer needs the 'd'). New code should use just 'qemu'. Yes, I know that changing it now means more merge conflicts to resolve later in the series. Oh well. > +static int qemudDomainDetachDeviceDiskLive(struct qemud_driver *driver, > + virDomainObjPtr vm, > + virDomainDeviceDefPtr dev, > + virBitmapPtr qemuCaps) > +{ > + virDomainDiskDefPtr disk = dev->data.disk; > + int ret = -1; > + > + switch (disk->device) { > + case VIR_DOMAIN_DISK_DEVICE_DISK: > + if (disk->bus == VIR_DOMAIN_DISK_BUS_VIRTIO) > + ret = qemuDomainDetachPciDiskDevice(driver, vm, dev, qemuCaps); > + else if (disk->bus == VIR_DOMAIN_DISK_BUS_SCSI) > + ret = qemuDomainDetachDiskDevice(driver, vm, dev, qemuCaps); > + else if (dev->data.disk->bus == VIR_DOMAIN_DISK_BUS_USB) > + ret = qemuDomainDetachDiskDevice(driver, vm, dev, qemuCaps); > + else > + qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", > + _("This type of disk cannot be hot unplugged")); > + break; > + default: > + break; Oops, no error message on that path. > + > +enum { > + QEMUD_DEVICE_ATTACH, QEMUD_DEVICE_DETACH, QEMUD_DEVICE_UPDATE, I flattened these out to one-per-line as part of renaming to QEMU_*. > + * update domain status forcibly because the domain status may be changed > * even if we attach the device failed. For example, a new controller may > * be created. > */ > - if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) > + if (!ret && virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) This change violates the comment right before it. I undid it. ACK with those nits fixed, so I pushed a modified version (shoot; I've lost an easy way to have git show my incremental diffs, due to the merge conflict resolution from the renames). -- Eric Blake eblake(a)redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org

2 1

Re: [libvirt] [PATCHv11 1/6] libvirt/qemu - Centralize device modification in the more flexible APIs
by Eric Blake 22 Apr '11

22 Apr '11

On 04/21/2011 01:22 AM, KAMEZAWA Hiroyuki wrote: > > Centralize device modification in the more flexible APIs, to allow future > honoring of additional flags. Explicitly reject the > VIR_DOMAIN_DEVICE_MODIFY_FORCE flag on attach/detach. > > Based on Eric Blake<eblake(a)redhat.com>'s work. > > From: Eric Blake <eblake(a)redhat.com> > Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu(a)jp.fujitsu.com> > > * src/qemu/qemu_driver.c > (qemudDomainAttachDevice)(qemudDomainAttachDeviceFlags): Swap bodies,rename... Where's the rename? > (qemudDomainDetachDevice, qemudDomainDetachDeviceFlags): Likewise. > --- > src/qemu/qemu_driver.c | 53 ++++++++++++++++++++++++++--------------------- > 1 files changed, 29 insertions(+), 24 deletions(-) > > diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c > index f6e503a..a8f3849 100644 > --- a/src/qemu/qemu_driver.c > +++ b/src/qemu/qemu_driver.c > @@ -3786,8 +3786,8 @@ cleanup: > } > > > -static int qemudDomainAttachDevice(virDomainPtr dom, > - const char *xml) > +static int qemudDomainAttachDeviceFlags(virDomainPtr dom, const char *xml, > + unsigned int flags) I adjusted this to do s/qemud/qemu/ for the lines of code actually touched in this patch, plus fallout. ACK and pushed with this squashed in. diff --git c/src/qemu/qemu_driver.c w/src/qemu/qemu_driver.c index a8f3849..1f5a899 100644 --- c/src/qemu/qemu_driver.c +++ w/src/qemu/qemu_driver.c @@ -3786,8 +3786,8 @@ cleanup: } -static int qemudDomainAttachDeviceFlags(virDomainPtr dom, const char *xml, - unsigned int flags) +static int qemuDomainAttachDeviceFlags(virDomainPtr dom, const char *xml, + unsigned int flags) { struct qemud_driver *driver = dom->conn->privateData; virDomainObjPtr vm; @@ -3951,10 +3951,10 @@ cleanup: return ret; } -static int qemudDomainAttachDevice(virDomainPtr dom, const char *xml) +static int qemuDomainAttachDevice(virDomainPtr dom, const char *xml) { - return qemudDomainAttachDeviceFlags(dom, xml, - VIR_DOMAIN_DEVICE_MODIFY_LIVE); + return qemuDomainAttachDeviceFlags(dom, xml, + VIR_DOMAIN_DEVICE_MODIFY_LIVE); } @@ -4080,8 +4080,8 @@ cleanup: } -static int qemudDomainDetachDeviceFlags(virDomainPtr dom, const char *xml, - unsigned int flags) +static int qemuDomainDetachDeviceFlags(virDomainPtr dom, const char *xml, + unsigned int flags) { struct qemud_driver *driver = dom->conn->privateData; virDomainObjPtr vm; @@ -4175,10 +4175,10 @@ cleanup: return ret; } -static int qemudDomainDetachDevice(virDomainPtr dom, const char *xml) +static int qemuDomainDetachDevice(virDomainPtr dom, const char *xml) { - return qemudDomainDetachDeviceFlags(dom, xml, - VIR_DOMAIN_DEVICE_MODIFY_LIVE); + return qemuDomainDetachDeviceFlags(dom, xml, + VIR_DOMAIN_DEVICE_MODIFY_LIVE); } static int qemudDomainGetAutostart(virDomainPtr dom, @@ -6980,10 +6980,10 @@ static virDriver qemuDriver = { qemudDomainStartWithFlags, /* domainCreateWithFlags */ qemudDomainDefine, /* domainDefineXML */ qemudDomainUndefine, /* domainUndefine */ - qemudDomainAttachDevice, /* domainAttachDevice */ - qemudDomainAttachDeviceFlags, /* domainAttachDeviceFlags */ - qemudDomainDetachDevice, /* domainDetachDevice */ - qemudDomainDetachDeviceFlags, /* domainDetachDeviceFlags */ + qemuDomainAttachDevice, /* domainAttachDevice */ + qemuDomainAttachDeviceFlags, /* domainAttachDeviceFlags */ + qemuDomainDetachDevice, /* domainDetachDevice */ + qemuDomainDetachDeviceFlags, /* domainDetachDeviceFlags */ qemuDomainUpdateDeviceFlags, /* domainUpdateDeviceFlags */ qemudDomainGetAutostart, /* domainGetAutostart */ qemudDomainSetAutostart, /* domainSetAutostart */ -- Eric Blake eblake(a)redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org

2 1

Re: [libvirt] [PATCHv11 0/6] libvirt/qemu - persistent modification of devices
by Eric Blake 22 Apr '11

22 Apr '11

I'm assuming you meant for this to go to libvir-list rather than libvirt-users; all my subsequent mails will swap the list reply. On 04/21/2011 01:19 AM, KAMEZAWA Hiroyuki wrote: > > Here is v11. Fixed comments/bugs and updated against the latest libvirt.git. > > Changes v10->v11: > - fixed comments on each patches > - fixed cgroup handling in patch 3. > - fixed MODIFY_CURRENT handling in patch 4. > > most of diff comes from refactoring qemu/qemu_driver.c > -- > conf/domain_conf.c | 40 ++ > conf/domain_conf.h | 5 > libvirt_private.syms | 3 > qemu/qemu_driver.c | 727 +++++++++++++++++++++++++++++++-------------------- > 4 files changed, 500 insertions(+), 275 deletions(-) > > > Thanks, > -Kame > > -- Eric Blake eblake(a)redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org

2 1

[libvirt] [PATCH 00/16] Implement migration v3 protocol
by Daniel P. Berrange 22 Apr '11

22 Apr '11

This is an major update to http://www.redhat.com/archives/libvir-list/2011-February/msg00257.html In this update - Bug fixes from previous review - Remove assert() usage - Introduce SPICE graphics relocation - Refactoring of peer2peer & tunnelled migration code - Support for v3 migration with p2p & tunnelled migration - A (broken) attempt to make job signal/status updates work with tunnelled migration The latter item seemed to be working fine, but I've since found that it can cause QEMU to deadlock. While we're sending QEMU a montor command to update status, QEMU may be trying to write some migration data and it often blocks on write() to the migration socket. This could be a bug with us not giving it a non-blocking socket, or it could be a QEMU limitation. To be investigated....

2 19

[libvirt] [PATCH] Fix QEMU tunnelled migration FD handling
by Daniel P. Berrange 21 Apr '11

21 Apr '11

The two ends of the pipe used for feeding QEMU tunnelled migration data were interchanged, so QEMU got given the "write" end instead of the "read" end. The qemuMigrationPrepareTunnel method was also immediately closing the "write" end of the pipe, so the stream failed to actually write anything. * src/qemu/qemu_migration.c: Swap tunnelled migration pipe FDs & don't close pipe given to stream --- src/qemu/qemu_migration.c | 7 ++++--- 1 files changed, 4 insertions(+), 3 deletions(-) diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c index bba76d5..7f4b111 100644 --- a/src/qemu/qemu_migration.c +++ b/src/qemu/qemu_migration.c @@ -301,7 +301,7 @@ qemuMigrationPrepareTunnel(struct qemud_driver *driver, vm->def->id = -1; if (pipe(dataFD) < 0 || - virSetCloseExec(dataFD[0]) < 0) { + virSetCloseExec(dataFD[1]) < 0) { virReportSystemError(errno, "%s", _("cannot create pipe for tunnelled migration")); goto endjob; @@ -318,7 +318,7 @@ qemuMigrationPrepareTunnel(struct qemud_driver *driver, /* Start the QEMU daemon, with the same command-line arguments plus * -incoming stdio (which qemu_command might convert to exec:cat or fd:n) */ - internalret = qemuProcessStart(dconn, driver, vm, "stdio", true, dataFD[1], + internalret = qemuProcessStart(dconn, driver, vm, "stdio", true, dataFD[0], NULL, VIR_VM_OP_MIGRATE_IN_START); if (internalret < 0) { qemuAuditDomainStart(vm, "migrated", false); @@ -332,7 +332,7 @@ qemuMigrationPrepareTunnel(struct qemud_driver *driver, goto endjob; } - if (virFDStreamOpen(st, dataFD[0]) < 0) { + if (virFDStreamOpen(st, dataFD[1]) < 0) { qemuAuditDomainStart(vm, "migrated", false); qemuProcessStop(driver, vm, 0); if (!vm->persistent) { @@ -344,6 +344,7 @@ qemuMigrationPrepareTunnel(struct qemud_driver *driver, _("cannot pass pipe for tunnelled migration")); goto endjob; } + dataFD[1] = -1; /* 'st' owns the FD now & will close it */ qemuAuditDomainStart(vm, "migrated", true); -- 1.7.4.4

2 1

[libvirt] [PATCH] Remove artificial minimum limit for guest memory
by Jiri Denemark 21 Apr '11

21 Apr '11

Remove the artificial minimum of 4096 KB for guest memory. It's drivers' job to set the limit if needed. --- src/libvirt.c | 6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-) diff --git a/src/libvirt.c b/src/libvirt.c index 9e6784b..10c3cdf 100644 --- a/src/libvirt.c +++ b/src/libvirt.c @@ -2738,7 +2738,7 @@ virDomainSetMaxMemory(virDomainPtr domain, unsigned long memory) virLibDomainError(VIR_ERR_OPERATION_DENIED, __FUNCTION__); goto error; } - if (memory < 4096) { + if (!memory) { virLibDomainError(VIR_ERR_INVALID_ARG, __FUNCTION__); goto error; } @@ -2792,7 +2792,7 @@ virDomainSetMemory(virDomainPtr domain, unsigned long memory) virLibDomainError(VIR_ERR_OPERATION_DENIED, __FUNCTION__); goto error; } - if (memory < 4096) { + if (!memory) { virLibDomainError(VIR_ERR_INVALID_ARG, __FUNCTION__); goto error; } @@ -2861,7 +2861,7 @@ virDomainSetMemoryFlags(virDomainPtr domain, unsigned long memory, goto error; } - if (memory < 4096) { + if (!memory) { virLibDomainError(VIR_ERR_INVALID_ARG, __FUNCTION__); goto error; } -- 1.7.5.rc1

2 2

[libvirt] tunnelled migration with libvirt 0.9.0
by Jason Krieg 21 Apr '11

21 Apr '11

Hi, is tunnelled migration with libvirt 0.9.0 and qemu-kvm 0.14.0 broken ? all I get is an error "unknown migration protocol" while looking at the code in src/qemu/qemu_migration.c src/qemu/qemu_migration.c: /* Start the QEMU daemon, with the same command-line arguments plus * -incoming stdin (which qemu_command might convert to exec:cat or fd:n) */ internalret = qemuProcessStart(dconn, driver, vm, "stdin", true, dataFD[1], NULL, VIR_VM_OP_MIGRATE_IN_START); migrateFrom is set to "stdin" but in src/qemu/qemu_command.c there is no such match which ends in the above error. Regards, Jason

5 7

[libvirt] [PATCH] PHYP: Adding reboot domain function
by Eduardo Otubo 21 Apr '11

21 Apr '11

Adding reboot <domain> function for pHyp driver. --- src/phyp/phyp_driver.c | 35 ++++++++++++++++++++++++++++++++++- 1 files changed, 34 insertions(+), 1 deletions(-) diff --git a/src/phyp/phyp_driver.c b/src/phyp/phyp_driver.c index bb0e0ac..228751d 100644 --- a/src/phyp/phyp_driver.c +++ b/src/phyp/phyp_driver.c @@ -3384,6 +3384,39 @@ cleanup: } static int +phypDomainReboot(virDomainPtr dom) +{ + int result = -1; + ConnectionData *connection_data = dom->conn->networkPrivateData; + virConnectPtr conn = dom->conn; + LIBSSH2_SESSION *session = connection_data->session; + phyp_driverPtr phyp_driver = conn->privateData; + int system_type = phyp_driver->system_type; + char *managed_system = phyp_driver->managed_system; + int exit_status = 0; + char *ret = NULL; + virBuffer buf = VIR_BUFFER_INITIALIZER; + + virBufferAddLit(&buf, "chsysstate"); + if (system_type == HMC) + virBufferVSprintf(&buf, " -m %s", managed_system); + virBufferVSprintf(&buf, + " -r lpar -o shutdown --id %d --immed --restart", + dom->id); + ret = phypExecBuffer(session, &buf, &exit_status, dom->conn, false); + + if (exit_status < 0) + goto cleanup; + + result = 0; + + cleanup: + VIR_FREE(ret); + + return result; +} + +static int phypDomainShutdown(virDomainPtr dom) { int result = -1; @@ -3707,7 +3740,7 @@ static virDriver phypDriver = { NULL, /* domainSuspend */ phypDomainResume, /* domainResume */ phypDomainShutdown, /* domainShutdown */ - NULL, /* domainReboot */ + phypDomainReboot, /* domainReboot */ phypDomainDestroy, /* domainDestroy */ NULL, /* domainGetOSType */ NULL, /* domainGetMaxMemory */ -- 1.7.1

2 1