- Devel - libvirt List Archives

[libvirt] [PATCH] libvirt-publican copyright tags
by David Jorm 25 Jan '10

25 Jan '10

I don't have commiter access on this. Can someone else please push it or give me access? --- a/en-US/Legal_Notice.xml +++ b/en-US/Legal_Notice.xml @@ -3,10 +3,10 @@ ]> <legalnotice> <para> - Copyright <trademark class="copyright"></trademark> &YEAR; &HOLDER; + Copyright <trademark class="copyright"></trademark> &YEAR; &HOLDER; </para> <para> - Copyright © 2009 Red Hat, Inc. and others. + Copyright © 2010 Red Hat, Inc. and others. </para> <para> The text of and illustrations in this document are licensed by Red Hat

3 3

[libvirt] esx: networkPtr not initialized
by Nimal I 24 Jan '10

24 Jan '10

hy, I'm tring to add a new API for getting the network info from ESX. I found out that I dont get a valid pointer for the networkPtr while initializing.(whereas DomainPtr is initialozed and gettting a valid pointer on a successfull connect itself). Can somebody let me know, what to do to initialize this pointer and further add a API for a netconfig. Thanks in Advance, -- Nimal.I

2 1

[libvirt] PCI Passthrough
by Aaron Clausen 22 Jan '10

22 Jan '10

I'm trying once again to get PCI passthrough working (KVM 84 on Ubuntu 9.10), and I'm getting this error : LC_ALL=C PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin /usr/bin/kvm -S -M pc-0.11 -m 4096 -smp 4 -name mailserver -uuid 76a83471-e94a-3658-fa61-8eceaa74ffc2 -monitor unix:/var/run/libvirt/qemu/mailserver.monitor,server,nowait -localtime -boot c -drive file=,if=ide,media=cdrom,index=2 -drive file=/var/lib/libvirt/images/mailserver.img,if=virtio,index=0,boot=on -drive file=/var/lib/libvirt/images/mailserver-2.img,if=virtio,index=1 -net nic,macaddr=54:52:00:1b:b2:56,vlan=0,model=virtio,name=virtio.0 -net tap,fd=17,vlan=0,name=tap.0 -serial pty -parallel none -usb -usbdevice tablet -vnc 127.0.0.1:0 -k en-us -vga cirrus -pcidevice host=0a:01.0 char device redirected to /dev/pts/0 get_real_device: /sys/bus/pci/devices/0000:0a:01.0/config: Permission denied init_assigned_device: Error: Couldn't get real device (0a:01.0)! Failed to initialize assigned device host=0a:01.0 Any thoughts? -- Aaron Clausen mightymartianca(a)gmail.com

1 0

[libvirt] [PATCH] qemu: Fix race between device rebind and kvm cleanup
by Chris Lalancette 22 Jan '10

22 Jan '10

Certain hypervisors (like qemu/kvm) map the PCI bar(s) on the host when doing device passthrough. This can lead to a race condition where the hypervisor is still cleaning up the device while libvirt is trying to re-attach it to the host device driver. To avoid this situation, we look through /proc/iomem, and if the hypervisor is still holding onto the bar (denoted by the string in the matcher variable), then we can wait around a bit for that to clear up. v2: Thanks to review by DV, make sure we wait the full timeout per-device Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/libvirt_private.syms | 1 + src/qemu/qemu_driver.c | 19 ++++++--- src/util/pci.c | 97 ++++++++++++++++++++++++++++++++++++++++++++++ src/util/pci.h | 1 + 4 files changed, 112 insertions(+), 6 deletions(-) diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms index c912d81..e42c090 100644 --- a/src/libvirt_private.syms +++ b/src/libvirt_private.syms @@ -439,6 +439,7 @@ pciGetDevice; pciFreeDevice; pciDettachDevice; pciReAttachDevice; +pciWaitForDeviceCleanup; pciResetDevice; pciDeviceSetManaged; pciDeviceGetManaged; diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index e14ed8f..55550ef 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -2277,12 +2277,19 @@ qemuDomainReAttachHostDevices(virConnectPtr conn, for (i = 0; i < pciDeviceListCount(pcidevs); i++) { pciDevice *dev = pciDeviceListGet(pcidevs, i); - if (pciDeviceGetManaged(dev) && - pciReAttachDevice(NULL, dev) < 0) { - virErrorPtr err = virGetLastError(); - VIR_ERROR(_("Failed to re-attach PCI device: %s"), - err ? err->message : ""); - virResetError(err); + int retries = 100; + if (pciDeviceGetManaged(dev)) { + while (pciWaitForDeviceCleanup(dev, "kvm_assigned_device") + && retries) { + usleep(100*1000); + retries--; + } + if (pciReAttachDevice(NULL, dev) < 0) { + virErrorPtr err = virGetLastError(); + VIR_ERROR(_("Failed to re-attach PCI device: %s"), + err ? err->message : ""); + virResetError(err); + } } } diff --git a/src/util/pci.c b/src/util/pci.c index 0defbfe..09535bd 100644 --- a/src/util/pci.c +++ b/src/util/pci.c @@ -883,6 +883,103 @@ pciReAttachDevice(virConnectPtr conn, pciDevice *dev) return pciUnBindDeviceFromStub(conn, dev, driver); } +/* Certain hypervisors (like qemu/kvm) map the PCI bar(s) on + * the host when doing device passthrough. This can lead to a race + * condition where the hypervisor is still cleaning up the device while + * libvirt is trying to re-attach it to the host device driver. To avoid + * this situation, we look through /proc/iomem, and if the hypervisor is + * still holding onto the bar (denoted by the string in the matcher variable), + * then we can wait around a bit for that to clear up. + * + * A typical /proc/iomem looks like this (snipped for brevity): + * 00010000-0008efff : System RAM + * 0008f000-0008ffff : reserved + * ... + * 00100000-cc9fcfff : System RAM + * 00200000-00483d3b : Kernel code + * 00483d3c-005c88df : Kernel data + * cc9fd000-ccc71fff : ACPI Non-volatile Storage + * ... + * d0200000-d02fffff : PCI Bus #05 + * d0200000-d021ffff : 0000:05:00.0 + * d0200000-d021ffff : e1000e + * d0220000-d023ffff : 0000:05:00.0 + * d0220000-d023ffff : e1000e + * ... + * f0000000-f0003fff : 0000:00:1b.0 + * f0000000-f0003fff : kvm_assigned_device + * + * Returns 0 if we are clear to continue, and 1 if the hypervisor is still + * holding onto the resource. + */ +int +pciWaitForDeviceCleanup(pciDevice *dev, const char *matcher) +{ + FILE *fp; + char line[160]; + unsigned long long start, end; + int consumed; + char *rest; + unsigned long long domain; + int bus, slot, function; + int in_matching_device; + int ret; + size_t match_depth; + + fp = fopen("/proc/iomem", "r"); + if (!fp) { + /* If we failed to open iomem, we just basically ignore the error. The + * unbind might succeed anyway, and besides, it's very likely we have + * no way to report the error + */ + VIR_DEBUG0("Failed to open /proc/iomem, trying to continue anyway"); + return 0; + } + + ret = 0; + in_matching_device = 0; + match_depth = 0; + while (fgets(line, sizeof(line), fp) != 0) { + /* the logic here is a bit confusing. For each line, we look to + * see if it matches the domain:bus:slot.function we were given. + * If this line matches the DBSF, then any subsequent lines indented + * by 2 spaces are the PCI regions for this device. It's also + * possible that none of the PCI regions are currently mapped, in + * which case we have no indented regions. This code handles all + * of these situations + */ + if (in_matching_device && (strspn(line, " ") == (match_depth + 2))) { + if (sscanf(line, "%Lx-%Lx : %n", &start, &end, &consumed) != 2) + continue; + + rest = line + consumed; + if (STRPREFIX(rest, matcher)) { + ret = 1; + break; + } + } + else { + in_matching_device = 0; + if (sscanf(line, "%Lx-%Lx : %n", &start, &end, &consumed) != 2) + continue; + + rest = line + consumed; + if (sscanf(rest, "%Lx:%x:%x.%x", &domain, &bus, &slot, &function) != 4) + continue; + + if (domain != dev->domain || bus != dev->bus || slot != dev->slot || + function != dev->function) + continue; + in_matching_device = 1; + match_depth = strspn(line, " "); + } + } + + fclose(fp); + + return ret; +} + static char * pciReadDeviceID(pciDevice *dev, const char *id_name) { diff --git a/src/util/pci.h b/src/util/pci.h index a1a19e7..e6ab137 100644 --- a/src/util/pci.h +++ b/src/util/pci.h @@ -81,5 +81,6 @@ int pciDeviceFileIterate(virConnectPtr conn, int pciDeviceIsAssignable(virConnectPtr conn, pciDevice *dev, int strict_acs_check); +int pciWaitForDeviceCleanup(pciDevice *dev, const char *matcher); #endif /* __VIR_PCI_H__ */ -- 1.6.6

2 2

[libvirt] [PATCH] qemu: Fix race between device rebind and kvm cleanup
by Chris Lalancette 22 Jan '10

22 Jan '10

Certain hypervisors (like qemu/kvm) map the PCI bar(s) on the host when doing device passthrough. This can lead to a race condition where the hypervisor is still cleaning up the device while libvirt is trying to re-attach it to the host device driver. To avoid this situation, we look through /proc/iomem, and if the hypervisor is still holding onto the bar (denoted by the string in the matcher variable), then we can wait around a bit for that to clear up. Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/libvirt_private.syms | 1 + src/qemu/qemu_driver.c | 19 ++++++--- src/util/pci.c | 97 ++++++++++++++++++++++++++++++++++++++++++++++ src/util/pci.h | 1 + 4 files changed, 112 insertions(+), 6 deletions(-) diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms index c912d81..e42c090 100644 --- a/src/libvirt_private.syms +++ b/src/libvirt_private.syms @@ -439,6 +439,7 @@ pciGetDevice; pciFreeDevice; pciDettachDevice; pciReAttachDevice; +pciWaitForDeviceCleanup; pciResetDevice; pciDeviceSetManaged; pciDeviceGetManaged; diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index e14ed8f..cde0034 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -2244,6 +2244,7 @@ qemuDomainReAttachHostDevices(virConnectPtr conn, { pciDeviceList *pcidevs; int i; + int retries = 100; if (!def->nhostdevs) return; @@ -2277,12 +2278,18 @@ qemuDomainReAttachHostDevices(virConnectPtr conn, for (i = 0; i < pciDeviceListCount(pcidevs); i++) { pciDevice *dev = pciDeviceListGet(pcidevs, i); - if (pciDeviceGetManaged(dev) && - pciReAttachDevice(NULL, dev) < 0) { - virErrorPtr err = virGetLastError(); - VIR_ERROR(_("Failed to re-attach PCI device: %s"), - err ? err->message : ""); - virResetError(err); + if (pciDeviceGetManaged(dev)) { + while (pciWaitForDeviceCleanup(dev, "kvm_assigned_device") + && retries) { + usleep(100*1000); + retries--; + } + if (pciReAttachDevice(NULL, dev) < 0) { + virErrorPtr err = virGetLastError(); + VIR_ERROR(_("Failed to re-attach PCI device: %s"), + err ? err->message : ""); + virResetError(err); + } } } diff --git a/src/util/pci.c b/src/util/pci.c index 0defbfe..09535bd 100644 --- a/src/util/pci.c +++ b/src/util/pci.c @@ -883,6 +883,103 @@ pciReAttachDevice(virConnectPtr conn, pciDevice *dev) return pciUnBindDeviceFromStub(conn, dev, driver); } +/* Certain hypervisors (like qemu/kvm) map the PCI bar(s) on + * the host when doing device passthrough. This can lead to a race + * condition where the hypervisor is still cleaning up the device while + * libvirt is trying to re-attach it to the host device driver. To avoid + * this situation, we look through /proc/iomem, and if the hypervisor is + * still holding onto the bar (denoted by the string in the matcher variable), + * then we can wait around a bit for that to clear up. + * + * A typical /proc/iomem looks like this (snipped for brevity): + * 00010000-0008efff : System RAM + * 0008f000-0008ffff : reserved + * ... + * 00100000-cc9fcfff : System RAM + * 00200000-00483d3b : Kernel code + * 00483d3c-005c88df : Kernel data + * cc9fd000-ccc71fff : ACPI Non-volatile Storage + * ... + * d0200000-d02fffff : PCI Bus #05 + * d0200000-d021ffff : 0000:05:00.0 + * d0200000-d021ffff : e1000e + * d0220000-d023ffff : 0000:05:00.0 + * d0220000-d023ffff : e1000e + * ... + * f0000000-f0003fff : 0000:00:1b.0 + * f0000000-f0003fff : kvm_assigned_device + * + * Returns 0 if we are clear to continue, and 1 if the hypervisor is still + * holding onto the resource. + */ +int +pciWaitForDeviceCleanup(pciDevice *dev, const char *matcher) +{ + FILE *fp; + char line[160]; + unsigned long long start, end; + int consumed; + char *rest; + unsigned long long domain; + int bus, slot, function; + int in_matching_device; + int ret; + size_t match_depth; + + fp = fopen("/proc/iomem", "r"); + if (!fp) { + /* If we failed to open iomem, we just basically ignore the error. The + * unbind might succeed anyway, and besides, it's very likely we have + * no way to report the error + */ + VIR_DEBUG0("Failed to open /proc/iomem, trying to continue anyway"); + return 0; + } + + ret = 0; + in_matching_device = 0; + match_depth = 0; + while (fgets(line, sizeof(line), fp) != 0) { + /* the logic here is a bit confusing. For each line, we look to + * see if it matches the domain:bus:slot.function we were given. + * If this line matches the DBSF, then any subsequent lines indented + * by 2 spaces are the PCI regions for this device. It's also + * possible that none of the PCI regions are currently mapped, in + * which case we have no indented regions. This code handles all + * of these situations + */ + if (in_matching_device && (strspn(line, " ") == (match_depth + 2))) { + if (sscanf(line, "%Lx-%Lx : %n", &start, &end, &consumed) != 2) + continue; + + rest = line + consumed; + if (STRPREFIX(rest, matcher)) { + ret = 1; + break; + } + } + else { + in_matching_device = 0; + if (sscanf(line, "%Lx-%Lx : %n", &start, &end, &consumed) != 2) + continue; + + rest = line + consumed; + if (sscanf(rest, "%Lx:%x:%x.%x", &domain, &bus, &slot, &function) != 4) + continue; + + if (domain != dev->domain || bus != dev->bus || slot != dev->slot || + function != dev->function) + continue; + in_matching_device = 1; + match_depth = strspn(line, " "); + } + } + + fclose(fp); + + return ret; +} + static char * pciReadDeviceID(pciDevice *dev, const char *id_name) { diff --git a/src/util/pci.h b/src/util/pci.h index a1a19e7..e6ab137 100644 --- a/src/util/pci.h +++ b/src/util/pci.h @@ -81,5 +81,6 @@ int pciDeviceFileIterate(virConnectPtr conn, int pciDeviceIsAssignable(virConnectPtr conn, pciDevice *dev, int strict_acs_check); +int pciWaitForDeviceCleanup(pciDevice *dev, const char *matcher); #endif /* __VIR_PCI_H__ */ -- 1.6.6

2 2

[libvirt] [PATCH] Add a rule to check for uses of readlink.
by Chris Lalancette 22 Jan '10

22 Jan '10

Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- .x-sc_prohibit_readlink | 2 ++ cfg.mk | 5 +++++ 2 files changed, 7 insertions(+), 0 deletions(-) create mode 100644 .x-sc_prohibit_readlink diff --git a/.x-sc_prohibit_readlink b/.x-sc_prohibit_readlink new file mode 100644 index 0000000..e7acb03 --- /dev/null +++ b/.x-sc_prohibit_readlink @@ -0,0 +1,2 @@ +^src/util/util\.c$ +^ChangeLog-old$ diff --git a/cfg.mk b/cfg.mk index 0f2d2a6..5013e4b 100644 --- a/cfg.mk +++ b/cfg.mk @@ -94,6 +94,11 @@ sc_prohibit_strncpy: msg='use virStrncpy, not strncpy' \ $(_prohibit_regexp) +sc_prohibit_readlink: + @re='readlink *\(' \ + msg='use virFileResolveLink, not readlink' \ + $(_prohibit_regexp) + sc_prohibit_gethostname: @re='gethostname *\(' \ msg='use virGetHostname, not gethostname' \ -- 1.6.6

3 3

[libvirt] [PATCH] Use virFileResolveLink instead of readlink in AppArmor
by Chris Lalancette 22 Jan '10

22 Jan '10

Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/security/security_apparmor.c | 17 +++++++++-------- 1 files changed, 9 insertions(+), 8 deletions(-) diff --git a/src/security/security_apparmor.c b/src/security/security_apparmor.c index f288645..0857d58 100644 --- a/src/security/security_apparmor.c +++ b/src/security/security_apparmor.c @@ -258,22 +258,23 @@ get_profile_name(virConnectPtr conn, virDomainObjPtr vm) static int use_apparmor(void) { - char libvirt_daemon[PATH_MAX]; int rc = -1; - ssize_t len = 0; + char *libvirt_daemon = NULL; - if ((len = readlink("/proc/self/exe", libvirt_daemon, - PATH_MAX - 1)) < 0) { + if (virFileResolveLink("/proc/self/exe", &libvirt_daemon) < 0) { virSecurityReportError(NULL, VIR_ERR_INTERNAL_ERROR, "%s", _("could not find libvirtd")); - return rc; + return -1; } - libvirt_daemon[len] = '\0'; if (access(APPARMOR_PROFILES_PATH, R_OK) != 0) - return rc; + goto cleanup; - return profile_status(libvirt_daemon, 1); + rc = profile_status(libvirt_daemon, 1); + +cleanup: + VIR_FREE(libvirt_daemon); + return rc; } /* Called on libvirtd startup to see if AppArmor is available */ -- 1.6.6

3 7

Re: [libvirt] [PATCH] Fix a compile warning in parthelper.c
by Chris Lalancette 22 Jan '10

22 Jan '10

On 01/22/2010 04:07 AM, Daniel Veillard wrote: > On Thu, Jan 21, 2010 at 11:33:18AM -0500, Chris Lalancette wrote: >> Signed-off-by: Chris Lalancette <clalance(a)redhat.com> >> --- >> src/storage/parthelper.c | 1 + >> 1 files changed, 1 insertions(+), 0 deletions(-) >> >> diff --git a/src/storage/parthelper.c b/src/storage/parthelper.c >> index 5df46e8..ab04842 100644 >> --- a/src/storage/parthelper.c >> +++ b/src/storage/parthelper.c >> @@ -34,6 +34,7 @@ >> >> #include <parted/parted.h> >> #include <stdio.h> >> +#include <string.h> >> >> /* we don't need to include the full internal.h just for this */ >> #define STREQ(a,b) (strcmp((a),(b)) == 0) > > ACK, Thanks, pushed. -- Chris Lalancette

1 0

[libvirt] [PATCH] Fix device assignment with root devices
by Chris Lalancette 22 Jan '10

22 Jan '10

The patches to add ACS checking to PCI device passthrough introduced a bug. With the current code, if you try to passthrough a device on the root bus (i.e. bus 0), then it denies the passthrough. This is because the code in pciDeviceIsBehindSwitchLackingACS() to check for a parent device doesn't take into account the possibility of the root bus. If we are on the root bus, it means we legitimately can't find a parent, and it also means that we don't have to worry about whether ACS is enabled. Therefore return 0 (indicating we don't lack ACS) from pciDeviceIsBehindSwitchLackingACS(). Signed-off-by: Chris Lalancette <clalance(a)redhat.com> --- src/util/pci.c | 19 ++++++++++++++----- 1 files changed, 14 insertions(+), 5 deletions(-) diff --git a/src/util/pci.c b/src/util/pci.c index 086b751..0defbfe 100644 --- a/src/util/pci.c +++ b/src/util/pci.c @@ -1198,11 +1198,20 @@ pciDeviceIsBehindSwitchLackingACS(virConnectPtr conn, { pciDevice *parent; - if (!(parent = pciGetParentDevice(conn, dev))) { - pciReportError(conn, VIR_ERR_NO_SUPPORT, - _("Failed to find parent device for %s"), - dev->name); - return -1; + parent = pciGetParentDevice(conn, dev); + if (!parent) { + /* if we have no parent, and this is the root bus, ACS doesn't come + * into play since devices on the root bus can't P2P without going + * through the root IOMMU. + */ + if (dev->bus == 0) + return 0; + else { + pciReportError(conn, VIR_ERR_NO_SUPPORT, + _("Failed to find parent device for %s"), + dev->name); + return -1; + } } /* XXX we should rather fail when we can't find device's parent and -- 1.6.6

2 2

[libvirt] [PATCH] Add virConnectGetVersion Python API
by Taizo ITO 22 Jan '10

22 Jan '10

Hi all, The attached patch adds a new python API call for retrieving the running hypervisor version used by a connection: virConnectGetVersion. -- Taizo ITO

2 1