- Devel - libvirt List Archives

[libvirt] [PATCH] Fix a crash when restarting libvirtd.
by Chris Lalancette 25 Jan '10

25 Jan '10

If you shutdown libvirtd while a domain with PCI devices is running, then try to restart libvirtd, libvirtd will crash. This happens because qemuUpdateActivePciHostdevs() is calling pciDeviceListSteal() with a dev of 0x0 (NULL), and then trying to dereference it. This patch fixes it up so that qemuUpdateActivePciHostdevs() steals the devices after first Get()'ting them, avoiding the crash. Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/qemu/qemu_driver.c | 6 ++++-- 1 files changed, 4 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 55550ef..bbdbe33 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -2147,6 +2147,7 @@ qemuUpdateActivePciHostdevs(struct qemud_driver *driver, virDomainDefPtr def) { pciDeviceList *pcidevs; + int i; int ret = -1; if (!def->nhostdevs) @@ -2155,8 +2156,9 @@ qemuUpdateActivePciHostdevs(struct qemud_driver *driver, if (!(pcidevs = qemuGetPciHostDeviceList(NULL, def))) return -1; - while (pciDeviceListCount(pcidevs) > 0) { - pciDevice *dev = pciDeviceListSteal(NULL, pcidevs, 0); + for (i = 0; i < pciDeviceListCount(pcidevs); i++) { + pciDevice *dev = pciDeviceListGet(pcidevs, i); + pciDeviceListSteal(NULL, pcidevs, dev); if (pciDeviceListAdd(NULL, driver->activePciHostdevs, dev) < 0) { -- 1.6.6

2 2

[libvirt] [PATCH] Revert "Fix libvirtd restart for domains with PCI passthrough devices"
by Chris Lalancette 25 Jan '10

25 Jan '10

This reverts commit cdc42d0a4865199a941d330dbb6ca1ef426323ae. As DanB pointed out, this patch is actually wrong. The real bug that was causing me to see this problem is a bug introduced in a RHEL-5 libvirt snapshot, and I'm going to fix the real bug there. Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/conf/domain_conf.c | 5 +++-- 1 files changed, 3 insertions(+), 2 deletions(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 6d0e2dc..e548d1d 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -2876,7 +2876,7 @@ static int virDomainHostdevSubsysPciDefParseXML(virConnectPtr conn, const xmlNodePtr node, virDomainHostdevDefPtr def, - int flags ATTRIBUTE_UNUSED) { + int flags) { int ret = -1; xmlNodePtr cur; @@ -2890,7 +2890,8 @@ virDomainHostdevSubsysPciDefParseXML(virConnectPtr conn, if (virDomainDevicePCIAddressParseXML(conn, cur, addr) < 0) goto out; - } else if (xmlStrEqual(cur->name, BAD_CAST "state")) { + } else if ((flags & VIR_DOMAIN_XML_INTERNAL_STATUS) && + xmlStrEqual(cur->name, BAD_CAST "state")) { /* Legacy back-compat. Don't add any more attributes here */ char *devaddr = virXMLPropString(cur, "devaddr"); if (devaddr && -- 1.6.6

2 2

[libvirt] [PATCH] qemuMonitorTextAttachDrive: avoid two leaks
by Jim Meyering 25 Jan '10

25 Jan '10

Coverity reported the first leak: return-(-1)-would-leak-"safe_str". I spotted the 2nd. >From 6326ab7b52f3550f24015c27407c1d3c2a1fd64d Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Mon, 25 Jan 2010 20:19:01 +0100 Subject: [PATCH] qemuMonitorTextAttachDrive: avoid two leaks * src/qemu/qemu_monitor_text.c (qemuMonitorTextAttachDrive): Most other failures in this function would "goto cleanup", but one mistakenly returned directly, skipping the cleanup and resulting in a leak. In addition, iterating the "try_command" loop would clobber, and thus leak, the "cmd" allocated on the first iteration, so be careful to free it in addition to "reply" beforehand. --- src/qemu/qemu_monitor_text.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_monitor_text.c b/src/qemu/qemu_monitor_text.c index ce5349b..380bcdc 100644 --- a/src/qemu/qemu_monitor_text.c +++ b/src/qemu/qemu_monitor_text.c @@ -1881,13 +1881,13 @@ try_command: if (qemudParseDriveAddReply(reply, driveAddr) < 0) { if (!tryOldSyntax && strstr(reply, "invalid char in expression")) { VIR_FREE(reply); + VIR_FREE(cmd); tryOldSyntax = 1; goto try_command; } qemudReportError (NULL, NULL, NULL, VIR_ERR_OPERATION_FAILED, _("adding %s disk failed: %s"), drivestr, reply); - VIR_FREE(reply); - return -1; + goto cleanup; } ret = 0; -- 1.7.0.rc0.127.gab8271

2 2

[libvirt] [PATCH] qemu: Search binaries in PATH instead of hardcoding /usr/bin
by Matthias Bolte 25 Jan '10

25 Jan '10

--- src/qemu/qemu_conf.c | 137 ++++++++++++++++++++++++++++++-------------------- 1 files changed, 82 insertions(+), 55 deletions(-) diff --git a/src/qemu/qemu_conf.c b/src/qemu/qemu_conf.c index 92a9348..6141ec6 100644 --- a/src/qemu/qemu_conf.c +++ b/src/qemu/qemu_conf.c @@ -379,20 +379,20 @@ static const struct qemu_feature_flags const arch_info_x86_64_flags [] = { /* The archicture tables for supported QEMU archs */ static const struct qemu_arch_info const arch_info_hvm[] = { - { "i686", 32, NULL, "/usr/bin/qemu", - "/usr/bin/qemu-system-x86_64", arch_info_i686_flags, 4 }, - { "x86_64", 64, NULL, "/usr/bin/qemu-system-x86_64", + { "i686", 32, NULL, "qemu", + "qemu-system-x86_64", arch_info_i686_flags, 4 }, + { "x86_64", 64, NULL, "qemu-system-x86_64", NULL, arch_info_x86_64_flags, 2 }, - { "arm", 32, NULL, "/usr/bin/qemu-system-arm", NULL, NULL, 0 }, - { "mips", 32, NULL, "/usr/bin/qemu-system-mips", NULL, NULL, 0 }, - { "mipsel", 32, NULL, "/usr/bin/qemu-system-mipsel", NULL, NULL, 0 }, - { "sparc", 32, NULL, "/usr/bin/qemu-system-sparc", NULL, NULL, 0 }, - { "ppc", 32, NULL, "/usr/bin/qemu-system-ppc", NULL, NULL, 0 }, + { "arm", 32, NULL, "qemu-system-arm", NULL, NULL, 0 }, + { "mips", 32, NULL, "qemu-system-mips", NULL, NULL, 0 }, + { "mipsel", 32, NULL, "qemu-system-mipsel", NULL, NULL, 0 }, + { "sparc", 32, NULL, "qemu-system-sparc", NULL, NULL, 0 }, + { "ppc", 32, NULL, "qemu-system-ppc", NULL, NULL, 0 }, }; static const struct qemu_arch_info const arch_info_xen[] = { - { "i686", 32, "xenner", "/usr/bin/xenner", NULL, arch_info_i686_flags, 4 }, - { "x86_64", 64, "xenner", "/usr/bin/xenner", NULL, arch_info_x86_64_flags, 2 }, + { "i686", 32, "xenner", "xenner", NULL, arch_info_i686_flags, 4 }, + { "x86_64", 64, "xenner", "xenner", NULL, arch_info_x86_64_flags, 2 }, }; @@ -768,8 +768,8 @@ qemudCapsInitGuest(virCapsPtr caps, int i; int haskvm = 0; int haskqemu = 0; - const char *kvmbin = NULL; - const char *binary = NULL; + char *kvmbin = NULL; + char *binary = NULL; time_t binary_mtime; virCapsGuestMachinePtr *machines = NULL; int nmachines = 0; @@ -779,10 +779,15 @@ qemudCapsInitGuest(virCapsPtr caps, /* Check for existance of base emulator, or alternate base * which can be used with magic cpu choice */ - if (access(info->binary, X_OK) == 0) - binary = info->binary; - else if (info->altbinary && access(info->altbinary, X_OK) == 0) - binary = info->altbinary; + binary = virFindFileInPath(info->binary); + + if (binary == NULL || access(binary, X_OK) != 0) { + VIR_FREE(binary); + binary = virFindFileInPath(info->altbinary); + + if (binary != NULL && access(binary, X_OK) != 0) + VIR_FREE(binary); + } /* Can use acceleration for KVM/KQEMU if * - host & guest arches match @@ -792,16 +797,30 @@ qemudCapsInitGuest(virCapsPtr caps, */ if (STREQ(info->arch, hostmachine) || (STREQ(hostmachine, "x86_64") && STREQ(info->arch, "i686"))) { - const char *const kvmbins[] = { "/usr/bin/qemu-kvm", /* Fedora */ - "/usr/bin/kvm" }; /* Upstream .spec */ + if (access("/dev/kvm", F_OK) == 0) { + const char *const kvmbins[] = { "qemu-kvm", /* Fedora */ + "kvm" }; /* Upstream .spec */ + + for (i = 0; i < ARRAY_CARDINALITY(kvmbins); ++i) { + kvmbin = virFindFileInPath(kvmbins[i]); + + if (kvmbin == NULL || access(kvmbin, X_OK) != 0) { + VIR_FREE(kvmbin); + continue; + } - for (i = 0; i < ARRAY_CARDINALITY(kvmbins); ++i) { - if (access(kvmbins[i], X_OK) == 0 && - access("/dev/kvm", F_OK) == 0) { haskvm = 1; - kvmbin = kvmbins[i]; - if (!binary) - binary = kvmbin; + + if (binary == NULL) { + binary = strdup(kvmbin); + + if (binary == NULL) { + virReportOOMError(NULL); + VIR_FREE(kvmbin); + return -1; + } + } + break; } } @@ -826,21 +845,18 @@ qemudCapsInitGuest(virCapsPtr caps, virCapsGuestMachinePtr machine; if (VIR_ALLOC(machine) < 0) { - virReportOOMError(NULL); - return -1; + goto no_memory; } if (!(machine->name = strdup(info->machine))) { - virReportOOMError(NULL); VIR_FREE(machine); - return -1; + goto no_memory; } if (VIR_ALLOC_N(machines, 1) < 0) { - virReportOOMError(NULL); VIR_FREE(machine->name); VIR_FREE(machine); - return -1; + goto no_memory; } machines[0] = machine; @@ -853,7 +869,7 @@ qemudCapsInitGuest(virCapsPtr caps, old_caps, &machines, &nmachines); if (probe && qemudProbeMachineTypes(binary, &machines, &nmachines) < 0) - return -1; + goto error; } /* We register kvm as the base emulator too, since we can @@ -865,21 +881,18 @@ qemudCapsInitGuest(virCapsPtr caps, binary, NULL, nmachines, - machines)) == NULL) { - for (i = 0; i < nmachines; i++) { - VIR_FREE(machines[i]->name); - VIR_FREE(machines[i]); - } - VIR_FREE(machines); - return -1; - } + machines)) == NULL) + goto error; + + machines = NULL; + nmachines = 0; guest->arch.defaultInfo.emulator_mtime = binary_mtime; if (qemudProbeCPUModels(binary, info->arch, &ncpus, NULL) == 0 && ncpus > 0 && !virCapabilitiesAddGuestFeature(guest, "cpuselection", 1, 0)) - return -1; + goto error; if (hvm) { if (virCapabilitiesAddGuestDomain(guest, @@ -888,7 +901,7 @@ qemudCapsInitGuest(virCapsPtr caps, NULL, 0, NULL) == NULL) - return -1; + goto error; if (haskqemu && virCapabilitiesAddGuestDomain(guest, @@ -897,7 +910,7 @@ qemudCapsInitGuest(virCapsPtr caps, NULL, 0, NULL) == NULL) - return -1; + goto error; if (haskvm) { virCapsGuestDomainPtr dom; @@ -911,9 +924,6 @@ qemudCapsInitGuest(virCapsPtr caps, binary_mtime = 0; } - machines = NULL; - nmachines = 0; - if (!STREQ(binary, kvmbin)) { int probe = 1; if (old_caps && binary_mtime) @@ -922,7 +932,7 @@ qemudCapsInitGuest(virCapsPtr caps, old_caps, &machines, &nmachines); if (probe && qemudProbeMachineTypes(kvmbin, &machines, &nmachines) < 0) - return -1; + goto error; } if ((dom = virCapabilitiesAddGuestDomain(guest, @@ -931,14 +941,12 @@ qemudCapsInitGuest(virCapsPtr caps, NULL, nmachines, machines)) == NULL) { - for (i = 0; i < nmachines; i++) { - VIR_FREE(machines[i]->name); - VIR_FREE(machines[i]); - } - VIR_FREE(machines); - return -1; + goto error; } + machines = NULL; + nmachines = 0; + dom->info.emulator_mtime = binary_mtime; } } else { @@ -948,7 +956,7 @@ qemudCapsInitGuest(virCapsPtr caps, NULL, 0, NULL) == NULL) - return -1; + goto error; } if (info->nflags) { @@ -957,11 +965,24 @@ qemudCapsInitGuest(virCapsPtr caps, info->flags[i].name, info->flags[i].default_on, info->flags[i].toggle) == NULL) - return -1; + goto error; } } + VIR_FREE(binary); + VIR_FREE(kvmbin); + return 0; + +no_memory: + virReportOOMError(NULL); + +error: + VIR_FREE(binary); + VIR_FREE(kvmbin); + virCapabilitiesFreeMachines(machines, nmachines); + + return -1; } @@ -1011,6 +1032,7 @@ virCapsPtr qemudCapsInit(virCapsPtr old_caps) { struct utsname utsname; virCapsPtr caps; int i; + char *xenner = NULL; /* Really, this never fails - look at the man-page. */ uname (&utsname); @@ -1051,7 +1073,9 @@ virCapsPtr qemudCapsInit(virCapsPtr old_caps) { goto no_memory; /* Then possibly the Xen paravirt guests (ie Xenner */ - if (access("/usr/bin/xenner", X_OK) == 0 && + xenner = virFindFileInPath("xenner"); + + if (xenner != NULL && access(xenner, X_OK) == 0 && access("/dev/kvm", F_OK) == 0) { for (i = 0 ; i < ARRAY_CARDINALITY(arch_info_xen) ; i++) /* Allow Xen 32-on-32, 32-on-64 and 64-on-64 */ @@ -1065,12 +1089,15 @@ virCapsPtr qemudCapsInit(virCapsPtr old_caps) { } } + VIR_FREE(xenner); + /* QEMU Requires an emulator in the XML */ virCapabilitiesSetEmulatorRequired(caps); return caps; no_memory: + VIR_FREE(xenner); virCapabilitiesFree(caps); return NULL; } -- 1.6.3.3

3 6

[libvirt] [PATCH] Look in /usr/libexec for the qemu-kvm binary.
by Chris Lalancette 25 Jan '10

25 Jan '10

On RHEL-5 the qemu-kvm binary is located in /usr/libexec. To reduce confusion for people trying to run upstream libvirt on RHEL-5 machines, make the qemu driver look in /usr/libexec for the qemu-kvm binary. Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/qemu/qemu_conf.c | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-) diff --git a/src/qemu/qemu_conf.c b/src/qemu/qemu_conf.c index c227fe1..2ba8366 100644 --- a/src/qemu/qemu_conf.c +++ b/src/qemu/qemu_conf.c @@ -797,7 +797,8 @@ qemudCapsInitGuest(virCapsPtr caps, */ if (STREQ(info->arch, hostmachine) || (STREQ(hostmachine, "x86_64") && STREQ(info->arch, "i686"))) { - const char *const kvmbins[] = { "/usr/bin/qemu-kvm", /* Fedora */ + const char *const kvmbins[] = { "/usr/libexec/qemu-kvm", /* RHEL */ + "/usr/bin/qemu-kvm", /* Fedora */ "/usr/bin/kvm" }; /* Upstream .spec */ for (i = 0; i < ARRAY_CARDINALITY(kvmbins); ++i) { -- 1.6.6

3 4

[libvirt] [PATCH] usbGetDevice: don't leak a "usbDevice" buffer on failure path
by Jim Meyering 25 Jan '10

25 Jan '10

Here's another leak fix, although this one is only on a failure path: >From a89551ecfcebfc3d45eff38923824a9590fe9f76 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Mon, 25 Jan 2010 16:44:13 +0100 Subject: [PATCH] usbGetDevice: don't leak a "usbDevice" buffer on failure path * src/util/hostusb.c (usbGetDevice): Free "dev" when returning NULL. --- src/util/hostusb.c | 4 +++- 1 files changed, 3 insertions(+), 1 deletions(-) diff --git a/src/util/hostusb.c b/src/util/hostusb.c index 9a37103..f635ce5 100644 --- a/src/util/hostusb.c +++ b/src/util/hostusb.c @@ -171,8 +171,10 @@ usbGetDevice(virConnectPtr conn, if (vendor) { /* Look up bus.dev by vendor:product */ - if (usbFindBusByVendor(conn, vendor, product, &bus, &devno) < 0) + if (usbFindBusByVendor(conn, vendor, product, &bus, &devno) < 0) { + VIR_FREE(dev); return NULL; + } } dev->bus = bus; -- 1.7.0.rc0.127.gab8271

2 2

[libvirt] [PATCH] qemuMonitorTextGetMemoryStats: plug a leak on an error path
by Jim Meyering 25 Jan '10

25 Jan '10

Coverity complained about a leak via this return -1 in qemu_monitor_text.c: int qemuMonitorTextGetMemoryStats(qemuMonitorPtr mon, virDomainMemoryStatPtr stats, unsigned int nr_stats) { char *reply = NULL; int ret = 0; char *offset; if (qemuMonitorCommand(mon, "info balloon", &reply) < 0) { qemudReportError(NULL, NULL, NULL, VIR_ERR_OPERATION_FAILED, "%s", _("could not query memory balloon statistics")); return -1; } That can happen because qemuMonitorCommand calls qemuMonitorCommandWithFd which calls qemuMonitorCommandWithHandler, which does this: 218 ret = qemuMonitorSend(mon, &msg); ... 228 if (msg.rxBuffer) { 229 *reply = msg.rxBuffer; 230 } else { 231 *reply = strdup(""); 232 if (!*reply) { 233 virReportOOMError(NULL); 234 return -1; 235 } 236 } 237 238 if (ret < 0) 239 virReportSystemError(NULL, msg.lastErrno, 240 _("cannot send monitor command '%s'"), cmd); 241 242 return ret; 243 } That function breaks contract by failing to free *reply when it returns a negative value. Here's the fix: >From 3b44df075f9d4330ec27d59eddaa0a32c20d7ac1 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Wed, 20 Jan 2010 18:24:47 +0100 Subject: [PATCH] qemuMonitorTextGetMemoryStats: plug a leak on an error path * src/qemu/qemu_monitor_text.c (qemuMonitorCommandWithHandler): Always free *reply, upon failure. --- src/qemu/qemu_monitor_text.c | 6 ++++-- 1 files changed, 4 insertions(+), 2 deletions(-) diff --git a/src/qemu/qemu_monitor_text.c b/src/qemu/qemu_monitor_text.c index c3848b5..d921c7e 100644 --- a/src/qemu/qemu_monitor_text.c +++ b/src/qemu/qemu_monitor_text.c @@ -1,7 +1,7 @@ /* * qemu_monitor_text.c: interaction with QEMU monitor console * - * Copyright (C) 2006-2009 Red Hat, Inc. + * Copyright (C) 2006-2010 Red Hat, Inc. * Copyright (C) 2006 Daniel P. Berrange * * This library is free software; you can redistribute it and/or @@ -235,9 +235,11 @@ qemuMonitorCommandWithHandler(qemuMonitorPtr mon, } } - if (ret < 0) + if (ret < 0) { virReportSystemError(NULL, msg.lastErrno, _("cannot send monitor command '%s'"), cmd); + VIR_FREE(*reply); + } return ret; } -- 1.6.6.516.gb72f

2 2

[libvirt] [PATCH 1/n] modernizing configure
by Eric Blake 25 Jan '10

25 Jan '10

I just built libvirt for the first time today, and noticed that configure fails on the first missing dependency, rather than trying harder to make it through and list all missing dependencies at the end. I ended up restarting configure multiple times. In the process, I also noticed that configure is somewhat out of date. So, what is the oldest version of autoconf and automake that libvirt will insist on supporting? For example, is it time to require automake 1.11.1 (the latest stable release) because of its security patch to avoid potential problems in 'make dist'? Or is it too soon to expect all distros to have an automake that new pre-installed, and more effort be put into supporting automake 1.9.6 and autoconf 2.59 (those being the implicit minimum requirements due to the use of gnulib)? Or somewhere in between? Meanwhile, how about this patch, for me getting my feet wet in helping to modernize the autotools usage, and hopefully leading up to future patches to make configure list all missing dependencies at once, rather than requiring multiple restarts.

4 10

[libvirt] [PATCH 0/4] Misc fixes to JSON/QMP mode for latest upstream QEMU
by Daniel P. Berrange 25 Jan '10

25 Jan '10

Just some misc fixes to make QMP mode work again for latest upstreqam QEMU GIT. NB this is OK because QMP isn't declared stable yet & isn't enabled in libvirt

2 10

[libvirt] [PATCH] Fix libvirtd restart for domains with PCI passthrough devices
by Chris Lalancette 25 Jan '10

25 Jan '10

When libvirtd shuts down, it places a <state/> tag in the XML state file it writes out for guests with PCI passthrough devices. For devices that are attached at bootup time, the state tag is empty. However, at libvirtd startup time, it ignores anything with a <state/> tag in the XML, effectively hiding the guest. I can think of at least 3 ways to fix this: 1) Don't throw an error on "unknown" tags in virDomainHostdevSubsysPciDefParseXML(). 2) Have virDomainLoadAllConfigs() pass the VIR_DOMAIN_XML_INTERNAL_STATUS flag when parsing the domain XML. 3) Remove the check for VIR_DOMAIN_XML_INTERNAL_STATUS when parsing the XML. I chose approach 3). My reasoning for this is that the <state> tag is a legitimate part of the XML, so we should always offer to parse it. This fixes the problem with reconnecting to domains that have PCI passthrough devices. Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/conf/domain_conf.c | 5 ++--- 1 files changed, 2 insertions(+), 3 deletions(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 74c2337..595c46c 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -2876,7 +2876,7 @@ static int virDomainHostdevSubsysPciDefParseXML(virConnectPtr conn, const xmlNodePtr node, virDomainHostdevDefPtr def, - int flags) { + int flags ATTRIBUTE_UNUSED) { int ret = -1; xmlNodePtr cur; @@ -2890,8 +2890,7 @@ virDomainHostdevSubsysPciDefParseXML(virConnectPtr conn, if (virDomainDevicePCIAddressParseXML(conn, cur, addr) < 0) goto out; - } else if ((flags & VIR_DOMAIN_XML_INTERNAL_STATUS) && - xmlStrEqual(cur->name, BAD_CAST "state")) { + } else if (xmlStrEqual(cur->name, BAD_CAST "state")) { /* Legacy back-compat. Don't add any more attributes here */ char *devaddr = virXMLPropString(cur, "devaddr"); if (devaddr && -- 1.6.6

3 3