[PATCH 1/2] virt-aa-helper: Avoid using RUNSTATEDIR
by Michal Privoznik
On some systems /run is mounted as:
tmpfs on /run type tmpfs (rw,nosuid,nodev,noexec,relatime,size=348508k,mode=755,inode64)
and /var/run is then just a symlink:
# ls -ld /var/run
lrwxrwxrwx 1 root root 4 Apr 23 2024 /var/run -> /run
But because we still think it's 2004 and FHS 2.3 is active we
have a rule in our meson.build which constructs RUNSTATEDIR as
the following:
runstatedir = get_option('runstatedir')
if runstatedir == ''
runstatedir = localstatedir / 'run'
endif
which (if unspecified on meson setup line) results in "/var/run".
This in turn means, when when we're generating an AppArmor
profile for a domain with allowed paths it contains stuff like:
/var/run/libvirt/qemu/swtpm/2-guest-swtpm.sock
But because of the aforementioned symlink the real path is:
/run/libvirt/qemu/swtpm/2-guest-swtpm.sock
and thus AppArmor denies access:
audit: type=1400 audit(1740480419.348:415): apparmor="DENIED" operation="connect" class="file" profile="libvirt-126f2720-6f8e-45ab-a886-ec9277079a67" name="/run/libvirt/qemu/swtpm/2-guest-swtpm.sock" pid=8080 comm="qemu-system-x86" requested_mask="wr" denied_mask="wr" fsuid=64055 ouid=64055
Fortunately, there's a nice trick: AppArmor profile variables. We
already use some of them (@{PROC}, @{HOME}, @{multiarch}) and
instead of RUNSTATEDIR we can use @{run} which is declared as:
# cat /etc/apparmor.d/tunables/run
@{run}=/run/ /var/run/
and thus covers both scenarios.
Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com>
---
src/security/virt-aa-helper.c | 16 ++++++++--------
1 file changed, 8 insertions(+), 8 deletions(-)
diff --git a/src/security/virt-aa-helper.c b/src/security/virt-aa-helper.c
index c255b64f35..eabc29e9cf 100644
--- a/src/security/virt-aa-helper.c
+++ b/src/security/virt-aa-helper.c
@@ -1238,8 +1238,8 @@ get_files(vahControl * ctl)
/* Unix socket for QEMU and swtpm to use */
virBufferAsprintf(&buf,
- " \"%s/libvirt/qemu/swtpm/%s-swtpm.sock\" rw,\n",
- RUNSTATEDIR, shortName);
+ " \"@{run}/libvirt/qemu/swtpm/%s-swtpm.sock\" rw,\n",
+ shortName);
/* Paths for swtpm to use: give it access to its state
* directory (state files and fsync on dir), log, and PID files.
*/
@@ -1253,8 +1253,8 @@ get_files(vahControl * ctl)
" \"%s/log/swtpm/libvirt/qemu/%s-swtpm.log\" w,\n",
LOCALSTATEDIR, ctl->def->name);
virBufferAsprintf(&buf,
- " \"%s/libvirt/qemu/swtpm/%s-swtpm.pid\" rw,\n",
- RUNSTATEDIR, shortName);
+ " \"@{run}/libvirt/qemu/swtpm/%s-swtpm.pid\" rw,\n",
+ shortName);
VIR_FREE(shortName);
}
@@ -1528,10 +1528,10 @@ main(int argc, char **argv)
LOCALSTATEDIR, ctl->def->name);
virBufferAsprintf(&buf, " \"%s/lib/libvirt/qemu/domain-%d-%.*s/*\" rw,\n",
LOCALSTATEDIR, ctl->def->id, 20, ctl->def->name);
- virBufferAsprintf(&buf, " \"%s/libvirt/**/%s.pid\" rwk,\n",
- RUNSTATEDIR, ctl->def->name);
- virBufferAsprintf(&buf, " \"%s/libvirt/**/*.tunnelmigrate.dest.%s\" rw,\n",
- RUNSTATEDIR, ctl->def->name);
+ virBufferAsprintf(&buf, " \"@{run}/libvirt/**/%s.pid\" rwk,\n",
+ ctl->def->name);
+ virBufferAsprintf(&buf, " \"@{run}/libvirt/**/*.tunnelmigrate.dest.%s\" rw,\n",
+ ctl->def->name);
}
if (ctl->files)
virBufferAdd(&buf, ctl->files, -1);
--
2.45.3
3 months, 2 weeks
[PATCH] virt-aa-helper: Avoid using RUNSTATEDIR
by Michal Privoznik
On some systems /run is mounted as:
tmpfs on /run type tmpfs (rw,nosuid,nodev,noexec,relatime,size=348508k,mode=755,inode64)
and /var/run is then just a symlink:
# ls -ld /var/run
lrwxrwxrwx 1 root root 4 Apr 23 2024 /var/run -> /run
But because we still think it's 2004 and FHS 2.3 is active we
have a rule in our meson.build which constructs RUNSTATEDIR as
the following:
runstatedir = get_option('runstatedir')
if runstatedir == ''
runstatedir = localstatedir / 'run'
endif
which (if unspecified on meson setup line) results in "/var/run".
This in turn means, when when we're generating an AppArmor
profile for a domain with allowed paths it contains stuff like:
/var/run/libvirt/qemu/swtpm/2-guest-swtpm.sock
But because of the aforementioned symlink the real path is:
/run/libvirt/qemu/swtpm/2-guest-swtpm.sock
and thus AppArmor denies access:
audit: type=1400 audit(1740480419.348:415): apparmor="DENIED" operation="connect" class="file" profile="libvirt-126f2720-6f8e-45ab-a886-ec9277079a67" name="/run/libvirt/qemu/swtpm/2-guest-swtpm.sock" pid=8080 comm="qemu-system-x86" requested_mask="wr" denied_mask="wr" fsuid=64055 ouid=64055
Fortunately, there's a nice trick: AppArmor profile variables. We
already use some of them (@{PROC}, @{HOME}, @{multiarch}) and
instead of RUNSTATEDIR we can use @{run} which is declared as:
# cat /etc/apparmor.d/tunables/run
@{run}=/run/ /var/run/
and thus covers both scenarios.
Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com>
---
src/security/virt-aa-helper.c | 16 ++++++++--------
1 file changed, 8 insertions(+), 8 deletions(-)
diff --git a/src/security/virt-aa-helper.c b/src/security/virt-aa-helper.c
index c255b64f35..eabc29e9cf 100644
--- a/src/security/virt-aa-helper.c
+++ b/src/security/virt-aa-helper.c
@@ -1238,8 +1238,8 @@ get_files(vahControl * ctl)
/* Unix socket for QEMU and swtpm to use */
virBufferAsprintf(&buf,
- " \"%s/libvirt/qemu/swtpm/%s-swtpm.sock\" rw,\n",
- RUNSTATEDIR, shortName);
+ " \"@{run}/libvirt/qemu/swtpm/%s-swtpm.sock\" rw,\n",
+ shortName);
/* Paths for swtpm to use: give it access to its state
* directory (state files and fsync on dir), log, and PID files.
*/
@@ -1253,8 +1253,8 @@ get_files(vahControl * ctl)
" \"%s/log/swtpm/libvirt/qemu/%s-swtpm.log\" w,\n",
LOCALSTATEDIR, ctl->def->name);
virBufferAsprintf(&buf,
- " \"%s/libvirt/qemu/swtpm/%s-swtpm.pid\" rw,\n",
- RUNSTATEDIR, shortName);
+ " \"@{run}/libvirt/qemu/swtpm/%s-swtpm.pid\" rw,\n",
+ shortName);
VIR_FREE(shortName);
}
@@ -1528,10 +1528,10 @@ main(int argc, char **argv)
LOCALSTATEDIR, ctl->def->name);
virBufferAsprintf(&buf, " \"%s/lib/libvirt/qemu/domain-%d-%.*s/*\" rw,\n",
LOCALSTATEDIR, ctl->def->id, 20, ctl->def->name);
- virBufferAsprintf(&buf, " \"%s/libvirt/**/%s.pid\" rwk,\n",
- RUNSTATEDIR, ctl->def->name);
- virBufferAsprintf(&buf, " \"%s/libvirt/**/*.tunnelmigrate.dest.%s\" rw,\n",
- RUNSTATEDIR, ctl->def->name);
+ virBufferAsprintf(&buf, " \"@{run}/libvirt/**/%s.pid\" rwk,\n",
+ ctl->def->name);
+ virBufferAsprintf(&buf, " \"@{run}/libvirt/**/*.tunnelmigrate.dest.%s\" rw,\n",
+ ctl->def->name);
}
if (ctl->files)
virBufferAdd(&buf, ctl->files, -1);
--
2.45.3
3 months, 2 weeks
[PATCH] qemu: Report disk bus as reported by agent in virDomainGetGuestInfo
by Peter Krempa
KubeVirt decided to report this to the users. In order to allow them to
use proper APIs expose the field as well.
Resolves: https://issues.redhat.com/browse/RHEL-80688
Signed-off-by: Peter Krempa <pkrempa(a)redhat.com>
---
docs/manpages/virsh.rst | 1 +
src/libvirt-domain.c | 1 +
src/qemu/qemu_driver.c | 9 +++++++++
3 files changed, 11 insertions(+)
diff --git a/docs/manpages/virsh.rst b/docs/manpages/virsh.rst
index f159c40631..baced15dec 100644
--- a/docs/manpages/virsh.rst
+++ b/docs/manpages/virsh.rst
@@ -2999,6 +2999,7 @@ returned:
* ``disk.<num>.serial`` - optional disk serial number
* ``disk.<num>.alias`` - the device alias of the disk (e.g. sda)
* ``disk.<num>.guest_alias`` - optional alias assigned to the disk
+* ``disk.<num>.guest_bus`` - bus type as reported by the guest
*--interface* returns:
* ``if.count`` - the number of interfaces defined on this domain
diff --git a/src/libvirt-domain.c b/src/libvirt-domain.c
index ae15ab1e5a..05101756a1 100644
--- a/src/libvirt-domain.c
+++ b/src/libvirt-domain.c
@@ -13273,6 +13273,7 @@ virDomainSetVcpu(virDomainPtr domain,
* "disk.<num>.alias" - the device alias of the disk (e.g. sda)
* "disk.<num>.guest_alias" - optional alias assigned to the disk, on Linux
* this is a name assigned by device mapper
+ * "disk.<num>.guest_bus" - disk bus as reported by the guest OS
*
* VIR_DOMAIN_GUEST_INFO_HOSTNAME:
* Returns information about the hostname of the domain. The typed
diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
index 3b8b2c0a1e..86b945d9b9 100644
--- a/src/qemu/qemu_driver.c
+++ b/src/qemu/qemu_driver.c
@@ -19278,6 +19278,15 @@ qemuAgentDiskInfoFormatParams(qemuAgentDiskInfo **info,
param_name, diskdef->dst) < 0)
return;
}
+
+ if (address->bus_type) {
+ g_snprintf(param_name, VIR_TYPED_PARAM_FIELD_LENGTH,
+ "disk.%zu.guest_bus", i);
+
+ if (virTypedParamsAddString(params, nparams, maxparams,
+ param_name, address->bus_type) < 0)
+ return;
+ }
}
if (info[i]->alias) {
--
2.48.1
3 months, 2 weeks
[libvirt PATCH] docs: formatdomain: fix typo in passt section
by Ján Tomko
A mismatch in backticks happened.
Fixes: a47a89a9d335c111a9c2fbb3f4e1c3a13001e74b
Reported-by: Yalan Zhang <yalzhang(a)redhat.com>
Signed-off-by: Ján Tomko <jtomko(a)redhat.com>
---
Pushed.
docs/formatdomain.rst | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/docs/formatdomain.rst b/docs/formatdomain.rst
index fdb3c2459e..cbe378e61d 100644
--- a/docs/formatdomain.rst
+++ b/docs/formatdomain.rst
@@ -5170,7 +5170,7 @@ values for both of these, or you can use the ``<ip>`` element
configure one IPv4 and one IPv6 address that passt's DHCP server can
provide to the guest.
-Unlike SLIRP, when no `<ip>`` address is specified, passt will by
+Unlike SLIRP, when no ``<ip>`` address is specified, passt will by
default provide the guest with an IP address, DNS server, etc. that
are identical to those settings on the host itself (through the magic
of the proxies and a separate network namespace, this doesn't create
--
2.48.1
3 months, 2 weeks
[PATCH 1/4] virpci: changed the work with PCI via libpciaccess
by Alexander Shursha
sysfs is used to get a list of PCI devices. It doesn't work under
FreeBSD. The libpciaccess library provides cross-platform functions
for accessing the PCI bus.
Signed-off-by: Alexander Shursha <kekek2(a)ya.ru>
---
src/meson.build | 1 +
src/util/virpci.c | 465 ++++++++++------------------------------
tests/qemuhotplugtest.c | 11 +-
tests/qemumemlocktest.c | 9 +
tests/qemuxmlconftest.c | 9 +
tests/virpcimock.c | 22 +-
6 files changed, 162 insertions(+), 355 deletions(-)
diff --git a/src/meson.build b/src/meson.build
index 9413192a55..39788ac4d7 100644
--- a/src/meson.build
+++ b/src/meson.build
@@ -411,6 +411,7 @@ libvirt_lib = shared_library(
dtrace_gen_objects,
dependencies: [
src_dep,
+ pciaccess_dep
],
link_args: libvirt_link_args,
link_whole: [
diff --git a/src/util/virpci.c b/src/util/virpci.c
index 90617e69c6..b5bbe73ece 100644
--- a/src/util/virpci.c
+++ b/src/util/virpci.c
@@ -2,6 +2,7 @@
* virpci.c: helper APIs for managing host PCI devices
*
* Copyright (C) 2009-2015 Red Hat, Inc.
+ * Copyright (C) 2024-2025 Future Crew, LLC
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
@@ -29,6 +30,7 @@
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
+#include <pciaccess.h>
#ifdef __linux__
# include <sys/utsname.h>
@@ -72,7 +74,7 @@ struct _virPCIDevice {
char *name; /* domain:bus:slot.function */
char id[PCI_ID_LEN]; /* product vendor */
- char *path;
+ struct pci_device *device;
/* The driver:domain which uses the device */
char *used_by_drvname;
@@ -359,121 +361,6 @@ virPCIDeviceGetCurrentDriverNameAndType(virPCIDevice *dev,
}
-static int
-virPCIDeviceConfigOpenInternal(virPCIDevice *dev, bool readonly, bool fatal)
-{
- int fd;
-
- fd = open(dev->path, readonly ? O_RDONLY : O_RDWR);
-
- if (fd < 0) {
- if (fatal) {
- virReportSystemError(errno,
- _("Failed to open config space file '%1$s'"),
- dev->path);
- } else {
- VIR_WARN("Failed to open config space file '%s': %s",
- dev->path, g_strerror(errno));
- }
- return -1;
- }
-
- VIR_DEBUG("%s %s: opened %s", dev->id, dev->name, dev->path);
- return fd;
-}
-
-static int
-virPCIDeviceConfigOpen(virPCIDevice *dev)
-{
- return virPCIDeviceConfigOpenInternal(dev, true, true);
-}
-
-static int
-virPCIDeviceConfigOpenTry(virPCIDevice *dev)
-{
- return virPCIDeviceConfigOpenInternal(dev, true, false);
-}
-
-static int
-virPCIDeviceConfigOpenWrite(virPCIDevice *dev)
-{
- return virPCIDeviceConfigOpenInternal(dev, false, true);
-}
-
-static void
-virPCIDeviceConfigClose(virPCIDevice *dev, int cfgfd)
-{
- if (VIR_CLOSE(cfgfd) < 0) {
- VIR_WARN("Failed to close config space file '%s': %s",
- dev->path, g_strerror(errno));
- }
-}
-
-
-static int
-virPCIDeviceRead(virPCIDevice *dev,
- int cfgfd,
- unsigned int pos,
- uint8_t *buf,
- unsigned int buflen)
-{
- memset(buf, 0, buflen);
- errno = 0;
-
- if (lseek(cfgfd, pos, SEEK_SET) != pos ||
- saferead(cfgfd, buf, buflen) != buflen) {
- VIR_DEBUG("Failed to read %u bytes at %u from '%s' : %s",
- buflen, pos, dev->path, g_strerror(errno));
- return -1;
- }
- return 0;
-}
-
-
-/**
- * virPCIDeviceReadN:
- * @dev: virPCIDevice object (used only to log name of config file)
- * @cfgfd: open file descriptor for device config file in sysfs
- * @pos: byte offset in the file to read from
- *
- * read "N" (where "N" is "8", "16", or "32", and appears at the end
- * of the function name) bytes from a PCI device's already-opened
- * sysfs config file and return them as the return value from the
- * function.
- *
- * Returns the value at @pos in the file, or 0 if there was an
- * error. NB: since 0 could be a valid value, occurrence of an error
- * must be determined by examining errno. errno is always reset to 0
- * before the seek/read is attempted (see virPCIDeviceRead()), so if
- * errno != 0 on return from one of these functions, then either the
- * seek or the read operation failed for some reason. If errno == 0
- * and the return value is 0, then the config file really does contain
- * the value 0 at @pos.
- */
-static uint8_t
-virPCIDeviceRead8(virPCIDevice *dev, int cfgfd, unsigned int pos)
-{
- uint8_t buf;
- virPCIDeviceRead(dev, cfgfd, pos, &buf, sizeof(buf));
- return buf;
-}
-
-static uint16_t
-virPCIDeviceRead16(virPCIDevice *dev, int cfgfd, unsigned int pos)
-{
- uint8_t buf[2];
- virPCIDeviceRead(dev, cfgfd, pos, &buf[0], sizeof(buf));
- return (buf[0] << 0) | (buf[1] << 8);
-}
-
-static uint32_t
-virPCIDeviceRead32(virPCIDevice *dev, int cfgfd, unsigned int pos)
-{
- uint8_t buf[4];
- virPCIDeviceRead(dev, cfgfd, pos, &buf[0], sizeof(buf));
- return (buf[0] << 0) | (buf[1] << 8) | (buf[2] << 16) | (buf[3] << 24);
-}
-
static int
virPCIDeviceReadClass(virPCIDevice *dev, uint16_t *device_class)
{
@@ -499,36 +386,6 @@ virPCIDeviceReadClass(virPCIDevice *dev, uint16_t *device_class)
return 0;
}
-static int
-virPCIDeviceWrite(virPCIDevice *dev,
- int cfgfd,
- unsigned int pos,
- uint8_t *buf,
- unsigned int buflen)
-{
- if (lseek(cfgfd, pos, SEEK_SET) != pos ||
- safewrite(cfgfd, buf, buflen) != buflen) {
- VIR_WARN("Failed to write to '%s' : %s", dev->path,
- g_strerror(errno));
- return -1;
- }
- return 0;
-}
-
-static void
-virPCIDeviceWrite16(virPCIDevice *dev, int cfgfd, unsigned int pos, uint16_t val)
-{
- uint8_t buf[2] = { (val >> 0), (val >> 8) };
- virPCIDeviceWrite(dev, cfgfd, pos, &buf[0], sizeof(buf));
-}
-
-static void
-virPCIDeviceWrite32(virPCIDevice *dev, int cfgfd, unsigned int pos, uint32_t val)
-{
- uint8_t buf[4] = { (val >> 0), (val >> 8), (val >> 16), (val >> 24) };
- virPCIDeviceWrite(dev, cfgfd, pos, &buf[0], sizeof(buf));
-}
-
typedef int (*virPCIDeviceIterPredicate)(virPCIDevice *, virPCIDevice *,
void *);
@@ -610,7 +467,6 @@ virPCIDeviceIterDevices(virPCIDeviceIterPredicate predicate,
*/
static int
virPCIDeviceFindCapabilityOffset(virPCIDevice *dev,
- int cfgfd,
unsigned int capability,
unsigned int *offset)
{
@@ -619,11 +475,13 @@ virPCIDeviceFindCapabilityOffset(virPCIDevice *dev,
*offset = 0; /* assume failure (*nothing* can be at offset 0) */
- status = virPCIDeviceRead16(dev, cfgfd, PCI_STATUS);
+ pci_device_cfg_read_u16(dev->device, &status, PCI_STATUS);
+
if (errno != 0 || !(status & PCI_STATUS_CAP_LIST))
goto error;
- pos = virPCIDeviceRead8(dev, cfgfd, PCI_CAPABILITY_LIST);
+ pci_device_cfg_read_u8(dev->device, &pos, PCI_CAPABILITY_LIST);
+
if (errno != 0)
goto error;
@@ -635,7 +493,9 @@ virPCIDeviceFindCapabilityOffset(virPCIDevice *dev,
* capabilities here.
*/
while (pos >= PCI_CONF_HEADER_LEN && pos != 0xff) {
- uint8_t capid = virPCIDeviceRead8(dev, cfgfd, pos);
+ uint8_t capid;
+ pci_device_cfg_read_u8(dev->device, &capid, pos);
+
if (errno != 0)
goto error;
@@ -646,7 +506,8 @@ virPCIDeviceFindCapabilityOffset(virPCIDevice *dev,
return 0;
}
- pos = virPCIDeviceRead8(dev, cfgfd, pos + 1);
+ pci_device_cfg_read_u8(dev->device, &pos, pos + 1);
+
if (errno != 0)
goto error;
}
@@ -665,7 +526,6 @@ virPCIDeviceFindCapabilityOffset(virPCIDevice *dev,
static unsigned int
virPCIDeviceFindExtendedCapabilityOffset(virPCIDevice *dev,
- int cfgfd,
unsigned int capability)
{
int ttl;
@@ -677,7 +537,7 @@ virPCIDeviceFindExtendedCapabilityOffset(virPCIDevice *dev,
pos = PCI_EXT_CAP_BASE;
while (ttl > 0 && pos >= PCI_EXT_CAP_BASE) {
- header = virPCIDeviceRead32(dev, cfgfd, pos);
+ header = pci_device_cfg_read_u32(dev->device, &header, pos);
if ((header & PCI_EXT_CAP_ID_MASK) == capability)
return pos;
@@ -693,7 +553,7 @@ virPCIDeviceFindExtendedCapabilityOffset(virPCIDevice *dev,
* not have FLR, 1 if it does, and -1 on error
*/
static bool
-virPCIDeviceDetectFunctionLevelReset(virPCIDevice *dev, int cfgfd)
+virPCIDeviceDetectFunctionLevelReset(virPCIDevice *dev)
{
uint32_t caps;
unsigned int pos;
@@ -707,7 +567,7 @@ virPCIDeviceDetectFunctionLevelReset(virPCIDevice *dev, int cfgfd)
* on SR-IOV NICs at the moment.
*/
if (dev->pcie_cap_pos) {
- caps = virPCIDeviceRead32(dev, cfgfd, dev->pcie_cap_pos + PCI_EXP_DEVCAP);
+ pci_device_cfg_read_u32(dev->device, &caps, dev->pcie_cap_pos + PCI_EXP_DEVCAP);
if (caps & PCI_EXP_DEVCAP_FLR) {
VIR_DEBUG("%s %s: detected PCIe FLR capability", dev->id, dev->name);
return true;
@@ -718,11 +578,13 @@ virPCIDeviceDetectFunctionLevelReset(virPCIDevice *dev, int cfgfd)
* the same thing, except for conventional PCI
* devices. This is not common yet.
*/
- if (virPCIDeviceFindCapabilityOffset(dev, cfgfd, PCI_CAP_ID_AF, &pos) < 0)
+ if (virPCIDeviceFindCapabilityOffset(dev, PCI_CAP_ID_AF, &pos) < 0)
goto error;
if (pos) {
- caps = virPCIDeviceRead16(dev, cfgfd, pos + PCI_AF_CAP);
+ uint16_t caps16;
+ pci_device_cfg_read_u16(dev->device, &caps16, pos + PCI_AF_CAP);
+ caps = caps16;
if (caps & PCI_AF_CAP_FLR) {
VIR_DEBUG("%s %s: detected PCI FLR capability", dev->id, dev->name);
return true;
@@ -754,13 +616,13 @@ virPCIDeviceDetectFunctionLevelReset(virPCIDevice *dev, int cfgfd)
* internal reset, not just a soft reset.
*/
static bool
-virPCIDeviceDetectPowerManagementReset(virPCIDevice *dev, int cfgfd)
+virPCIDeviceDetectPowerManagementReset(virPCIDevice *dev)
{
if (dev->pci_pm_cap_pos) {
uint32_t ctl;
/* require the NO_SOFT_RESET bit is clear */
- ctl = virPCIDeviceRead32(dev, cfgfd, dev->pci_pm_cap_pos + PCI_PM_CTRL);
+ pci_device_cfg_read_u32(dev->device, &ctl, dev->pci_pm_cap_pos + PCI_PM_CTRL);
if (!(ctl & PCI_PM_CTRL_NO_SOFT_RESET)) {
VIR_DEBUG("%s %s: detected PM reset capability", dev->id, dev->name);
return true;
@@ -811,26 +673,22 @@ virPCIDeviceIsParent(virPCIDevice *dev, virPCIDevice *check, void *data)
uint8_t header_type, secondary, subordinate;
virPCIDevice **best = data;
int ret = 0;
- int fd;
if (dev->address.domain != check->address.domain)
return 0;
- if ((fd = virPCIDeviceConfigOpenTry(check)) < 0)
- return 0;
-
/* Is it a bridge? */
ret = virPCIDeviceReadClass(check, &device_class);
if (ret < 0 || device_class != PCI_CLASS_BRIDGE_PCI)
- goto cleanup;
+ return ret;
/* Is it a plane? */
- header_type = virPCIDeviceRead8(check, fd, PCI_HEADER_TYPE);
+ pci_device_cfg_read_u8(check->device, &header_type, PCI_HEADER_TYPE);
if ((header_type & PCI_HEADER_TYPE_MASK) != PCI_HEADER_TYPE_BRIDGE)
- goto cleanup;
+ return ret;
- secondary = virPCIDeviceRead8(check, fd, PCI_SECONDARY_BUS);
- subordinate = virPCIDeviceRead8(check, fd, PCI_SUBORDINATE_BUS);
+ pci_device_cfg_read_u8(check->device, &secondary, PCI_SECONDARY_BUS);
+ pci_device_cfg_read_u8(check->device, &subordinate, PCI_SUBORDINATE_BUS);
VIR_DEBUG("%s %s: found parent device %s", dev->id, dev->name, check->name);
@@ -838,8 +696,7 @@ virPCIDeviceIsParent(virPCIDevice *dev, virPCIDevice *check, void *data)
* the direct parent. No further work is necessary
*/
if (dev->address.bus == secondary) {
- ret = 1;
- goto cleanup;
+ return 1;
}
/* otherwise, SRIOV allows VFs to be on different buses than their PFs.
@@ -850,35 +707,26 @@ virPCIDeviceIsParent(virPCIDevice *dev, virPCIDevice *check, void *data)
if (*best == NULL) {
*best = virPCIDeviceNew(&check->address);
if (*best == NULL) {
- ret = -1;
- goto cleanup;
+ return -1;
}
} else {
/* OK, we had already recorded a previous "best" match for the
* parent. See if the current device is more restrictive than the
* best, and if so, make it the new best
*/
- int bestfd;
uint8_t best_secondary;
- if ((bestfd = virPCIDeviceConfigOpenTry(*best)) < 0)
- goto cleanup;
- best_secondary = virPCIDeviceRead8(*best, bestfd, PCI_SECONDARY_BUS);
- virPCIDeviceConfigClose(*best, bestfd);
+ pci_device_cfg_read_u8((*best)->device, &best_secondary, PCI_SECONDARY_BUS);
if (secondary > best_secondary) {
virPCIDeviceFree(*best);
*best = virPCIDeviceNew(&check->address);
if (*best == NULL) {
- ret = -1;
- goto cleanup;
+ return -1;
}
}
}
}
-
- cleanup:
- virPCIDeviceConfigClose(check, fd);
return ret;
}
@@ -902,15 +750,14 @@ virPCIDeviceGetParent(virPCIDevice *dev, virPCIDevice **parent)
*/
static int
virPCIDeviceTrySecondaryBusReset(virPCIDevice *dev,
- int cfgfd,
virPCIDeviceList *inactiveDevs)
{
g_autoptr(virPCIDevice) parent = NULL;
g_autoptr(virPCIDevice) conflict = NULL;
uint8_t config_space[PCI_CONF_LEN];
uint16_t ctl;
- int ret = -1;
- int parentfd;
+ pciaddr_t bytes_read;
+ pciaddr_t bytes_written;
/* Refuse to do a secondary bus reset if there are other
* devices/functions behind the bus are used by the host
@@ -932,8 +779,6 @@ virPCIDeviceTrySecondaryBusReset(virPCIDevice *dev,
dev->name);
return -1;
}
- if ((parentfd = virPCIDeviceConfigOpenWrite(parent)) < 0)
- goto out;
VIR_DEBUG("%s %s: doing a secondary bus reset", dev->id, dev->name);
@@ -941,38 +786,37 @@ virPCIDeviceTrySecondaryBusReset(virPCIDevice *dev,
* for the supplied device since we refuse to do a reset if there
* are multiple devices/functions
*/
- if (virPCIDeviceRead(dev, cfgfd, 0, config_space, PCI_CONF_LEN) < 0) {
+ pci_device_cfg_read(dev->device, config_space, 0, PCI_CONF_LEN, &bytes_read);
+ if (bytes_read < PCI_CONF_LEN) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Failed to read PCI config space for %1$s"),
dev->name);
- goto out;
+ return -1;
}
/* Read the control register, set the reset flag, wait 200ms,
* unset the reset flag and wait 200ms.
*/
- ctl = virPCIDeviceRead16(dev, parentfd, PCI_BRIDGE_CONTROL);
- virPCIDeviceWrite16(parent, parentfd, PCI_BRIDGE_CONTROL,
- ctl | PCI_BRIDGE_CTL_RESET);
+ pci_device_cfg_read_u16(parent->device, &ctl, PCI_BRIDGE_CONTROL);
+
+ pci_device_cfg_write_u16(parent->device, ctl | PCI_BRIDGE_CTL_RESET, PCI_BRIDGE_CONTROL);
g_usleep(200 * 1000); /* sleep 200ms */
- virPCIDeviceWrite16(parent, parentfd, PCI_BRIDGE_CONTROL, ctl);
+ pci_device_cfg_write_u16(parent->device, ctl, PCI_BRIDGE_CONTROL);
g_usleep(200 * 1000); /* sleep 200ms */
- if (virPCIDeviceWrite(dev, cfgfd, 0, config_space, PCI_CONF_LEN) < 0) {
+ pci_device_cfg_write(dev->device, config_space, 0, PCI_CONF_LEN, &bytes_written);
+ if (bytes_written < PCI_CONF_LEN) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Failed to restore PCI config space for %1$s"),
dev->name);
- goto out;
+ return -1;
}
- ret = 0;
- out:
- virPCIDeviceConfigClose(parent, parentfd);
- return ret;
+ return 0;
}
/* Power management reset attempts to reset a device using a
@@ -980,16 +824,19 @@ virPCIDeviceTrySecondaryBusReset(virPCIDevice *dev,
* above we require the device supports a full internal reset.
*/
static int
-virPCIDeviceTryPowerManagementReset(virPCIDevice *dev, int cfgfd)
+virPCIDeviceTryPowerManagementReset(virPCIDevice *dev)
{
uint8_t config_space[PCI_CONF_LEN];
uint32_t ctl;
+ pciaddr_t bytes_read;
+ pciaddr_t bytes_written;
if (!dev->pci_pm_cap_pos)
return -1;
/* Save and restore the device's config space. */
- if (virPCIDeviceRead(dev, cfgfd, 0, &config_space[0], PCI_CONF_LEN) < 0) {
+ pci_device_cfg_read(dev->device, config_space, 0, PCI_CONF_LEN, &bytes_read);
+ if (bytes_read < PCI_CONF_LEN) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Failed to read PCI config space for %1$s"),
dev->name);
@@ -998,20 +845,19 @@ virPCIDeviceTryPowerManagementReset(virPCIDevice *dev, int cfgfd)
VIR_DEBUG("%s %s: doing a power management reset", dev->id, dev->name);
- ctl = virPCIDeviceRead32(dev, cfgfd, dev->pci_pm_cap_pos + PCI_PM_CTRL);
+ pci_device_cfg_read_u32(dev->device, &ctl, dev->pci_pm_cap_pos + PCI_PM_CTRL);
ctl &= ~PCI_PM_CTRL_STATE_MASK;
- virPCIDeviceWrite32(dev, cfgfd, dev->pci_pm_cap_pos + PCI_PM_CTRL,
- ctl | PCI_PM_CTRL_STATE_D3hot);
+ pci_device_cfg_write_u32(dev->device, ctl | PCI_PM_CTRL_STATE_D3hot, dev->pci_pm_cap_pos + PCI_PM_CTRL);
g_usleep(10 * 1000); /* sleep 10ms */
- virPCIDeviceWrite32(dev, cfgfd, dev->pci_pm_cap_pos + PCI_PM_CTRL,
- ctl | PCI_PM_CTRL_STATE_D0);
+ pci_device_cfg_write_u32(dev->device, ctl | PCI_PM_CTRL_STATE_D0, dev->pci_pm_cap_pos + PCI_PM_CTRL);
g_usleep(10 * 1000); /* sleep 10ms */
- if (virPCIDeviceWrite(dev, cfgfd, 0, &config_space[0], PCI_CONF_LEN) < 0) {
+ pci_device_cfg_write(dev->device, config_space, 0, PCI_CONF_LEN, &bytes_written);
+ if (bytes_written < PCI_CONF_LEN) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Failed to restore PCI config space for %1$s"),
dev->name);
@@ -1046,10 +892,10 @@ virPCIDeviceTryPowerManagementReset(virPCIDevice *dev, int cfgfd)
* Always returns success (0) (for now)
*/
static int
-virPCIDeviceInit(virPCIDevice *dev, int cfgfd)
+virPCIDeviceInit(virPCIDevice *dev)
{
dev->is_pcie = false;
- if (virPCIDeviceFindCapabilityOffset(dev, cfgfd, PCI_CAP_ID_EXP, &dev->pcie_cap_pos) < 0) {
+ if (virPCIDeviceFindCapabilityOffset(dev, PCI_CAP_ID_EXP, &dev->pcie_cap_pos) < 0) {
/* an unprivileged process is unable to read *all* of a
* device's PCI config (it can only read the first 64
* bytes, which isn't enough for see the Express
@@ -1065,18 +911,13 @@ virPCIDeviceInit(virPCIDevice *dev, int cfgfd)
* -1), then we blindly assume the most likely outcome -
* PCIe.
*/
- off_t configLen = virFileLength(virPCIDeviceGetConfigPath(dev), -1);
-
- if (configLen != 256)
- dev->is_pcie = true;
-
} else {
dev->is_pcie = (dev->pcie_cap_pos != 0);
}
- virPCIDeviceFindCapabilityOffset(dev, cfgfd, PCI_CAP_ID_PM, &dev->pci_pm_cap_pos);
- dev->has_flr = virPCIDeviceDetectFunctionLevelReset(dev, cfgfd);
- dev->has_pm_reset = virPCIDeviceDetectPowerManagementReset(dev, cfgfd);
+ virPCIDeviceFindCapabilityOffset(dev, PCI_CAP_ID_PM, &dev->pci_pm_cap_pos);
+ dev->has_flr = virPCIDeviceDetectFunctionLevelReset(dev);
+ dev->has_pm_reset = virPCIDeviceDetectPowerManagementReset(dev);
return 0;
}
@@ -1089,7 +930,6 @@ virPCIDeviceReset(virPCIDevice *dev,
g_autofree char *drvName = NULL;
virPCIStubDriver drvType;
int ret = -1;
- int fd = -1;
int hdrType = -1;
if (virPCIGetHeaderType(dev, &hdrType) < 0)
@@ -1114,29 +954,26 @@ virPCIDeviceReset(virPCIDevice *dev,
* be redundant.
*/
if (virPCIDeviceGetCurrentDriverNameAndType(dev, &drvName, &drvType) < 0)
- goto cleanup;
+ return -1;
if (drvType == VIR_PCI_STUB_DRIVER_VFIO) {
VIR_DEBUG("Device %s is bound to %s - skip reset", dev->name, drvName);
ret = 0;
- goto cleanup;
+ return 0;
}
VIR_DEBUG("Resetting device %s", dev->name);
- if ((fd = virPCIDeviceConfigOpenWrite(dev)) < 0)
- goto cleanup;
-
- if (virPCIDeviceInit(dev, fd) < 0)
- goto cleanup;
+ if (virPCIDeviceInit(dev) < 0)
+ return -1;
/* KVM will perform FLR when starting and stopping
* a guest, so there is no need for us to do it here.
*/
if (dev->has_flr) {
ret = 0;
- goto cleanup;
+ return 0;
}
/* If the device supports PCI power management reset,
@@ -1144,11 +981,11 @@ virPCIDeviceReset(virPCIDevice *dev,
* the function, not the whole device.
*/
if (dev->has_pm_reset)
- ret = virPCIDeviceTryPowerManagementReset(dev, fd);
+ ret = virPCIDeviceTryPowerManagementReset(dev);
/* Bus reset is not an option with the root bus */
if (ret < 0 && dev->address.bus != 0)
- ret = virPCIDeviceTrySecondaryBusReset(dev, fd, inactiveDevs);
+ ret = virPCIDeviceTrySecondaryBusReset(dev, inactiveDevs);
if (ret < 0) {
virErrorPtr err = virGetLastError();
@@ -1159,8 +996,6 @@ virPCIDeviceReset(virPCIDevice *dev,
_("no FLR, PM reset or bus reset available"));
}
- cleanup:
- virPCIDeviceConfigClose(dev, fd);
return ret;
}
@@ -1756,28 +1591,6 @@ virPCIDeviceReattach(virPCIDevice *dev,
return 0;
}
-static char *
-virPCIDeviceReadID(virPCIDevice *dev, const char *id_name)
-{
- g_autofree char *path = NULL;
- g_autofree char *id_str = NULL;
-
- path = virPCIFile(dev->name, id_name);
-
- /* ID string is '0xNNNN\n' ... i.e. 7 bytes */
- if (virFileReadAll(path, 7, &id_str) < 0)
- return NULL;
-
- /* Check for 0x suffix */
- if (id_str[0] != '0' || id_str[1] != 'x')
- return NULL;
-
- /* Chop off the newline; we know the string is 7 bytes */
- id_str[6] = '\0';
-
- return g_steal_pointer(&id_str);
-}
-
bool
virPCIDeviceAddressIsValid(virPCIDeviceAddress *addr,
bool report)
@@ -1865,9 +1678,9 @@ virPCIDeviceExists(const virPCIDeviceAddress *addr)
virPCIDevice *
virPCIDeviceNew(const virPCIDeviceAddress *address)
{
+ struct pci_device * device;
+
g_autoptr(virPCIDevice) dev = NULL;
- g_autofree char *vendor = NULL;
- g_autofree char *product = NULL;
dev = g_new0(virPCIDevice, 1);
@@ -1875,31 +1688,21 @@ virPCIDeviceNew(const virPCIDeviceAddress *address)
dev->name = virPCIDeviceAddressAsString(&dev->address);
- dev->path = g_strdup_printf(PCI_SYSFS "devices/%s/config", dev->name);
-
- if (!virFileExists(dev->path)) {
- virReportSystemError(errno,
- _("Device %1$s not found: could not access %2$s"),
- dev->name, dev->path);
+ pci_system_init();
+ device = pci_device_find_by_slot(address->domain, address->bus, address->slot, address->function);
+ if (!device)
return NULL;
- }
-
- vendor = virPCIDeviceReadID(dev, "vendor");
- product = virPCIDeviceReadID(dev, "device");
-
- if (!vendor || !product) {
- virReportError(VIR_ERR_INTERNAL_ERROR,
- _("Failed to read product/vendor ID for %1$s"),
- dev->name);
+ dev->device = g_memdup(device, sizeof(*dev->device));
+ if (!dev->device) {
+ virReportSystemError(errno,
+ _("Not found device domain: %1$d, bus: %2$d, slot: %3$d, function: %4$d"),
+ address->domain, address->bus, address->slot, address->function);
return NULL;
}
-
- /* strings contain '0x' prefix */
- if (g_snprintf(dev->id, sizeof(dev->id), "%s %s", &vendor[2],
- &product[2]) >= sizeof(dev->id)) {
+ if (g_snprintf(dev->id, sizeof(dev->id), "%x %x", dev->device->vendor_id, dev->device->device_id) >= sizeof(dev->id)) {
virReportError(VIR_ERR_INTERNAL_ERROR,
- _("dev->id buffer overflow: %1$s %2$s"),
- &vendor[2], &product[2]);
+ _("dev->id buffer overflow: %1$d %2$d"),
+ dev->device->vendor_id, dev->device->device_id);
return NULL;
}
@@ -1918,10 +1721,10 @@ virPCIDeviceCopy(virPCIDevice *dev)
/* shallow copy to take care of most attributes */
*copy = *dev;
- copy->path = NULL;
- copy->used_by_drvname = copy->used_by_domname = NULL;
+ copy->device = NULL;
+ copy->device = g_memdup(dev->device, sizeof(*dev->device));
+ copy->name = copy->used_by_drvname = copy->used_by_domname = copy->stubDriverName = NULL;
copy->name = g_strdup(dev->name);
- copy->path = g_strdup(dev->path);
copy->used_by_drvname = g_strdup(dev->used_by_drvname);
copy->used_by_domname = g_strdup(dev->used_by_domname);
copy->stubDriverName = g_strdup(dev->stubDriverName);
@@ -1936,10 +1739,11 @@ virPCIDeviceFree(virPCIDevice *dev)
return;
VIR_DEBUG("%s %s: freeing", dev->id, dev->name);
g_free(dev->name);
- g_free(dev->path);
g_free(dev->used_by_drvname);
g_free(dev->used_by_domname);
g_free(dev->stubDriverName);
+ if (dev->device)
+ g_free(dev->device);
g_free(dev);
}
@@ -1971,9 +1775,9 @@ virPCIDeviceGetName(virPCIDevice *dev)
* config file.
*/
const char *
-virPCIDeviceGetConfigPath(virPCIDevice *dev)
+virPCIDeviceGetConfigPath(virPCIDevice *dev G_GNUC_UNUSED)
{
- return dev->path;
+ return NULL;
}
void virPCIDeviceSetManaged(virPCIDevice *dev, bool managed)
@@ -2484,47 +2288,37 @@ virPCIDeviceDownstreamLacksACS(virPCIDevice *dev)
uint16_t flags;
uint16_t ctrl;
unsigned int pos;
- int fd;
- int ret = 0;
uint16_t device_class;
- if ((fd = virPCIDeviceConfigOpen(dev)) < 0)
+ if (virPCIDeviceInit(dev) < 0) {
return -1;
-
- if (virPCIDeviceInit(dev, fd) < 0) {
- ret = -1;
- goto cleanup;
}
if (virPCIDeviceReadClass(dev, &device_class) < 0)
- goto cleanup;
+ return 0;
pos = dev->pcie_cap_pos;
if (!pos || device_class != PCI_CLASS_BRIDGE_PCI)
- goto cleanup;
+ return 0;
- flags = virPCIDeviceRead16(dev, fd, pos + PCI_EXP_FLAGS);
+ pci_device_cfg_read_u16(dev->device, &flags, pos + PCI_EXP_FLAGS);
if (((flags & PCI_EXP_FLAGS_TYPE) >> 4) != PCI_EXP_TYPE_DOWNSTREAM)
- goto cleanup;
+ return 0;
- pos = virPCIDeviceFindExtendedCapabilityOffset(dev, fd, PCI_EXT_CAP_ID_ACS);
+ pos = virPCIDeviceFindExtendedCapabilityOffset(dev, PCI_EXT_CAP_ID_ACS);
if (!pos) {
VIR_DEBUG("%s %s: downstream port lacks ACS", dev->id, dev->name);
- ret = 1;
- goto cleanup;
+ return 1;
}
- ctrl = virPCIDeviceRead16(dev, fd, pos + PCI_EXT_ACS_CTRL);
+ pci_device_cfg_read_u16(dev->device, &ctrl, pos + PCI_EXT_ACS_CTRL);
if ((ctrl & PCI_EXT_CAP_ACS_ENABLED) != PCI_EXT_CAP_ACS_ENABLED) {
VIR_DEBUG("%s %s: downstream port has ACS disabled",
dev->id, dev->name);
- ret = 1;
- goto cleanup;
+ return 1;
}
- cleanup:
- virPCIDeviceConfigClose(dev, fd);
- return ret;
+ return 0;
}
static int
@@ -3189,48 +2983,27 @@ virPCIDeviceGetVPD(virPCIDevice *dev G_GNUC_UNUSED)
int
virPCIDeviceIsPCIExpress(virPCIDevice *dev)
{
- int fd;
- int ret = -1;
-
- if ((fd = virPCIDeviceConfigOpen(dev)) < 0)
- return ret;
-
- if (virPCIDeviceInit(dev, fd) < 0)
- goto cleanup;
-
- ret = dev->is_pcie;
+ if (virPCIDeviceInit(dev) < 0)
+ return -1;
- cleanup:
- virPCIDeviceConfigClose(dev, fd);
- return ret;
+ return dev->is_pcie;
}
int
virPCIDeviceHasPCIExpressLink(virPCIDevice *dev)
{
- int fd;
- int ret = -1;
uint16_t cap, type;
-
- if ((fd = virPCIDeviceConfigOpen(dev)) < 0)
- return ret;
-
- if (virPCIDeviceInit(dev, fd) < 0)
- goto cleanup;
+ if (virPCIDeviceInit(dev) < 0)
+ return -1;
if (dev->pcie_cap_pos == 0) {
- ret = 0;
- goto cleanup;
+ return 0;
}
- cap = virPCIDeviceRead16(dev, fd, dev->pcie_cap_pos + PCI_CAP_FLAGS);
+ pci_device_cfg_read_u16(dev->device, &cap, dev->pcie_cap_pos + PCI_CAP_FLAGS);
type = (cap & PCI_EXP_FLAGS_TYPE) >> 4;
- ret = type != PCI_EXP_TYPE_ROOT_INT_EP && type != PCI_EXP_TYPE_ROOT_EC;
-
- cleanup:
- virPCIDeviceConfigClose(dev, fd);
- return ret;
+ return type != PCI_EXP_TYPE_ROOT_INT_EP && type != PCI_EXP_TYPE_ROOT_EC;
}
int
@@ -3242,53 +3015,39 @@ virPCIDeviceGetLinkCapSta(virPCIDevice *dev,
unsigned int *sta_width)
{
uint32_t t;
- int fd;
- int ret = -1;
-
- if ((fd = virPCIDeviceConfigOpen(dev)) < 0)
- return ret;
-
- if (virPCIDeviceInit(dev, fd) < 0)
- goto cleanup;
+ uint16_t t16;
+ if (virPCIDeviceInit(dev) < 0)
+ return -1;
if (!dev->pcie_cap_pos) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("pci device %1$s is not a PCI-Express device"),
dev->name);
- goto cleanup;
+ return -1;
}
- t = virPCIDeviceRead32(dev, fd, dev->pcie_cap_pos + PCI_EXP_LNKCAP);
+ pci_device_cfg_read_u32(dev->device, &t, dev->pcie_cap_pos + PCI_EXP_LNKCAP);
*cap_port = t >> 24;
*cap_speed = t & PCI_EXP_LNKCAP_SPEED;
*cap_width = (t & PCI_EXP_LNKCAP_WIDTH) >> 4;
- t = virPCIDeviceRead16(dev, fd, dev->pcie_cap_pos + PCI_EXP_LNKSTA);
+ pci_device_cfg_read_u16(dev->device, &t16, dev->pcie_cap_pos + PCI_EXP_LNKSTA);
+ t = t16;
*sta_speed = t & PCI_EXP_LNKSTA_SPEED;
*sta_width = (t & PCI_EXP_LNKSTA_WIDTH) >> 4;
- ret = 0;
-
- cleanup:
- virPCIDeviceConfigClose(dev, fd);
- return ret;
+ return 0;
}
int virPCIGetHeaderType(virPCIDevice *dev, int *hdrType)
{
- int fd;
uint8_t type;
*hdrType = -1;
- if ((fd = virPCIDeviceConfigOpen(dev)) < 0)
- return -1;
-
- type = virPCIDeviceRead8(dev, fd, PCI_HEADER_TYPE);
-
- virPCIDeviceConfigClose(dev, fd);
+ pci_device_cfg_read_u8(dev->device, &type, PCI_HEADER_TYPE);
type &= PCI_HEADER_TYPE_MASK;
if (type >= VIR_PCI_HEADER_LAST) {
diff --git a/tests/qemuhotplugtest.c b/tests/qemuhotplugtest.c
index d2a1f5acf1..fd8339bd30 100644
--- a/tests/qemuhotplugtest.c
+++ b/tests/qemuhotplugtest.c
@@ -18,12 +18,14 @@
*/
#include <config.h>
+#include "testutils.h"
+
+#ifdef __linux__
#include "qemu/qemu_alias.h"
#include "qemu/qemu_conf.h"
#include "qemu/qemu_hotplug.h"
#include "qemumonitortestutils.h"
-#include "testutils.h"
#include "testutilsqemu.h"
#include "testutilsqemuschema.h"
#include "virhostdev.h"
@@ -816,3 +818,10 @@ VIR_TEST_MAIN_PRELOAD(mymain,
VIR_TEST_MOCK("domaincaps"),
VIR_TEST_MOCK("virprocess"),
VIR_TEST_MOCK("qemuhotplug"));
+#else
+int
+main(void)
+{
+ return EXIT_AM_SKIP;
+}
+#endif
diff --git a/tests/qemumemlocktest.c b/tests/qemumemlocktest.c
index a2888732f7..5f62c80665 100644
--- a/tests/qemumemlocktest.c
+++ b/tests/qemumemlocktest.c
@@ -7,6 +7,8 @@
#include "testutils.h"
+#ifdef __linux__
+
#ifdef WITH_QEMU
# include "internal.h"
@@ -137,3 +139,10 @@ main(void)
}
#endif /* WITH_QEMU */
+#else
+int
+main(void)
+{
+ return EXIT_AM_SKIP;
+}
+#endif
diff --git a/tests/qemuxmlconftest.c b/tests/qemuxmlconftest.c
index 1f0068864a..6f84812109 100644
--- a/tests/qemuxmlconftest.c
+++ b/tests/qemuxmlconftest.c
@@ -7,6 +7,8 @@
#include "testutils.h"
+#ifdef __linux__
+
#ifdef WITH_QEMU
# include "internal.h"
@@ -3051,3 +3053,10 @@ int main(void)
}
#endif /* WITH_QEMU */
+#else
+int
+main(void)
+{
+ return EXIT_AM_SKIP;
+}
+#endif
diff --git a/tests/virpcimock.c b/tests/virpcimock.c
index 5b923c63ce..36bb57edb0 100644
--- a/tests/virpcimock.c
+++ b/tests/virpcimock.c
@@ -1,5 +1,6 @@
/*
* Copyright (C) 2013 Red Hat, Inc.
+ * Copyright (C) 2024-2025 Future Crew, LLC
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
@@ -22,7 +23,7 @@
#include "virpcivpdpriv.h"
-#if defined(__linux__) || defined(__FreeBSD__) || defined(__APPLE__)
+#if defined(__linux__) || defined(__APPLE__)
# define VIR_MOCK_LOOKUP_MAIN
# include "virmock.h"
# include "virpci.h"
@@ -42,6 +43,10 @@ static int (*real___open_2)(const char *path, int flags);
static int (*real_close)(int fd);
static DIR * (*real_opendir)(const char *name);
static char *(*real_virFileCanonicalizePath)(const char *path);
+static int (*real_scandir)(const char *restrict dirp,
+ struct dirent ***restrict namelist,
+ typeof(int(const struct dirent *)) *filter,
+ typeof(int(const struct dirent **, const struct dirent **)) *compar);
static char *fakerootdir;
@@ -955,6 +960,7 @@ init_syms(void)
VIR_MOCK_REAL_INIT(opendir);
# endif
VIR_MOCK_REAL_INIT(virFileCanonicalizePath);
+ VIR_MOCK_REAL_INIT(scandir);
}
static void
@@ -1172,6 +1178,20 @@ virFileCanonicalizePath(const char *path)
return real_virFileCanonicalizePath(newpath);
}
+int scandir(const char *restrict dirp, struct dirent ***restrict namelist,
+ typeof(int(const struct dirent *)) *filter,
+ typeof(int(const struct dirent **, const struct dirent **)) *compar)
+{
+ g_autofree char *newpath = NULL;
+
+ init_syms();
+
+ if (getrealpath(&newpath, dirp) < 0)
+ return -1;
+
+ return real_scandir(newpath, namelist, filter, compar);
+}
+
# include "virmockstathelpers.c"
#else
--
2.47.1
3 months, 2 weeks
[PATCH v2 0/4] Add support for getting load averages from guest agent
by Martin Kletzander
I'm drawing blanks when it comes to commit messages today. But this still has
time, because even though the patches are in QEMU upstream, but not a part of a
release yet. However nothing should break if we add support earlier...
Right? O:-)
v2:
- split patches
- add it to guestinfo instead of domstats
- some more macro deduplication
Martin Kletzander (4):
Add load average information type into virDomainGetGuestInfo
qemu_agent: Add qemuAgentGetLoadAvg()
qemu: Add support for VIR_DOMAIN_GUEST_INFO_LOAD
virsh: Add support for VIR_DOMAIN_GUEST_INFO_LOAD
docs/manpages/virsh.rst | 8 ++++-
include/libvirt/libvirt-domain.h | 1 +
src/libvirt-domain.c | 8 +++++
src/qemu/qemu_agent.c | 55 ++++++++++++++++++++++++++++++++
src/qemu/qemu_agent.h | 6 ++++
src/qemu/qemu_driver.c | 21 +++++++++++-
tests/qemuagenttest.c | 53 ++++++++++++++++++++++++++++++
tools/virsh-domain.c | 6 ++++
8 files changed, 156 insertions(+), 2 deletions(-)
--
2.48.1
3 months, 2 weeks
[libvirt PATCH 0/3] ci: patches to make integration tests work
by Pavel Hrdina
I've tested this on our hosts that are using for gitlab integration
tests and everything seems to be working now.
There were multiple issues I've encountered:
- on Fedora 41 current selinux policy silently blocks virtnwfilterd
to create_socket_perms for packet_socket resulting in this error:
internal error: setup of pcap handle failed: can't mmap rx ring: Permission denied
I've created temporary selinux module to allow this permission, need
to post a patch to fedora selinux policy to fix this.
- libvirt-tck uses for some tests sleep(30) to wait for VM to boot
but on our hosts it sometimes took a bit longer resulting in
random failures, posted patches to libvirt-tck to fix this issue
- lcitool is not able to prepare Fedora 41 vm as there is no
python3-libdnf5 by default installed and ansible fails with error
Pavel Hrdina (3):
ci: use iptables to run libvirt-tck
ci: use Fedora 41 to run integration tests
ci: refresh with 'lcitool manifest'
ci/buildenv/{alpine-319.sh => alpine-321.sh} | 0
...e-319.Dockerfile => alpine-321.Dockerfile} | 2 +-
ci/gitlab/builds.yml | 8 ++++----
ci/gitlab/containers.yml | 4 ++--
ci/integration.yml | 20 +++++++++----------
ci/jobs.sh | 3 +++
ci/manifest.yml | 2 +-
7 files changed, 21 insertions(+), 18 deletions(-)
rename ci/buildenv/{alpine-319.sh => alpine-321.sh} (100%)
rename ci/containers/{alpine-319.Dockerfile => alpine-321.Dockerfile} (98%)
--
2.48.1
3 months, 2 weeks
[PATCH V2 0/4] qemu: Small cleanups to SaveImageGetCompressionProgram
by Jim Fehlig
V2 of
https://lists.libvirt.org/archives/list/devel@lists.libvirt.org/message/S...
Changes in V2:
* V1 moved the special-case handling of invalid dump image format to
doCoreDump. V2 removes it, and fails on invalid configuration similar
to the other save-related operations.
* New patch4 that moves checking for a valid save image format from
time of use to config file parsing at driver startup
Jim Fehlig (4):
qemu: Move declaration of virQEMUSaveFormat to header file
qemu: Don't ignore dump image format errors
qemu: Change return value of SaveImageGetCompressionProgram
qemu: Check for valid save image formats when loading driver config
src/qemu/qemu_conf.c | 35 ++++++++++++---
src/qemu/qemu_conf.h | 6 +--
src/qemu/qemu_driver.c | 30 ++++---------
src/qemu/qemu_saveimage.c | 92 +++++++--------------------------------
src/qemu/qemu_saveimage.h | 25 +++++++++--
src/qemu/qemu_snapshot.c | 8 ++--
6 files changed, 80 insertions(+), 116 deletions(-)
--
2.43.0
3 months, 2 weeks
[PATCH] qemu: Add guest load average to domain stats
by Martin Kletzander
Since this is already in QEMU we can add this stat which might be useful
for people.
Resolves: https://issues.redhat.com/browse/RHEL-71883
Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com>
---
docs/manpages/virsh.rst | 3 +++
src/libvirt-domain.c | 3 +++
src/qemu/qemu_agent.c | 59 +++++++++++++++++++++++++++++++++++++++++
src/qemu/qemu_agent.h | 6 +++++
src/qemu/qemu_driver.c | 32 ++++++++++++++++++++++
tests/qemuagenttest.c | 53 ++++++++++++++++++++++++++++++++++++
6 files changed, 156 insertions(+)
diff --git a/docs/manpages/virsh.rst b/docs/manpages/virsh.rst
index 06c2802b3f9f..60b556cb0cd0 100644
--- a/docs/manpages/virsh.rst
+++ b/docs/manpages/virsh.rst
@@ -2408,6 +2408,9 @@ When selecting the *--state* group the following fields are returned:
for bank <index> in cache monitor <num>
* ``cpu.cache.monitor.<num>.bank.<index>.bytes`` - the number of bytes
of last level cache that the domain is using on cache bank <index>
+* ``cpu.load.1m`` - average load in guest for last 1 minute
+* ``cpu.load.5m`` - average load in guest for last 5 minutes
+* ``cpu.load.15m`` - average load in guest for last 15 minutes
*--balloon* returns:
diff --git a/src/libvirt-domain.c b/src/libvirt-domain.c
index 072cc3225579..35f5e926e52b 100644
--- a/src/libvirt-domain.c
+++ b/src/libvirt-domain.c
@@ -12272,6 +12272,9 @@ virConnectGetDomainCapabilities(virConnectPtr conn,
* last level cache that the
* domain is using on cache
* bank <index>
+ * "cpu.load.1m" - average load in guest for last 1 minute
+ * "cpu.load.5m" - average load in guest for last 5 minutes
+ * "cpu.load.15m" - average load in guest for last 15 minutes
*
* VIR_DOMAIN_STATS_BALLOON:
* Return memory balloon device information.
diff --git a/src/qemu/qemu_agent.c b/src/qemu/qemu_agent.c
index 43fca86f10ed..6dbe8c52d738 100644
--- a/src/qemu/qemu_agent.c
+++ b/src/qemu/qemu_agent.c
@@ -2568,3 +2568,62 @@ int qemuAgentGetDisks(qemuAgent *agent,
g_free(*disks);
return -1;
}
+
+
+int
+qemuAgentGetLoadAvg(qemuAgent *agent,
+ double *load1m,
+ double *load5m,
+ double *load15m,
+ bool report_unsupported)
+{
+ g_autoptr(virJSONValue) cmd = NULL;
+ g_autoptr(virJSONValue) reply = NULL;
+ virJSONValue *data = NULL;
+ int rc;
+
+ if (load1m)
+ *load1m = 0;
+
+ if (load5m)
+ *load5m = 0;
+
+ if (load15m)
+ *load15m = 0;
+
+ if (!(cmd = qemuAgentMakeCommand("guest-get-load", NULL)))
+ return -1;
+
+ if ((rc = qemuAgentCommandFull(agent, cmd, &reply, agent->timeout,
+ report_unsupported)) < 0)
+ return rc;
+
+ if (!(data = virJSONValueObjectGetObject(reply, "return"))) {
+ virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+ _("qemu agent didn't return an array of loads"));
+ return -1;
+ }
+
+ if (load1m &&
+ virJSONValueObjectGetNumberDouble(data, "load1m", load1m) < 0) {
+ virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+ _("'load1m' missing in reply of guest-get-load"));
+ return -1;
+ }
+
+ if (load5m &&
+ virJSONValueObjectGetNumberDouble(data, "load5m", load5m) < 0) {
+ virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+ _("'load5m' missing in reply of guest-get-load"));
+ return -1;
+ }
+
+ if (load15m &&
+ virJSONValueObjectGetNumberDouble(data, "load15m", load15m) < 0) {
+ virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+ _("'load15m' missing in reply of guest-get-load"));
+ return -1;
+ }
+
+ return 0;
+}
diff --git a/src/qemu/qemu_agent.h b/src/qemu/qemu_agent.h
index f98586e8f8ab..cd17a98d3924 100644
--- a/src/qemu/qemu_agent.h
+++ b/src/qemu/qemu_agent.h
@@ -195,3 +195,9 @@ int qemuAgentSSHRemoveAuthorizedKeys(qemuAgent *agent,
int qemuAgentGetDisks(qemuAgent *mon,
qemuAgentDiskInfo ***disks,
bool report_unsupported);
+
+int qemuAgentGetLoadAvg(qemuAgent *agent,
+ double *load1m,
+ double *load5m,
+ double *load15m,
+ bool report_unsupported);
diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
index 80c918312b8f..d9279b3a5804 100644
--- a/src/qemu/qemu_driver.c
+++ b/src/qemu/qemu_driver.c
@@ -16937,6 +16937,36 @@ qemuDomainGetStatsCpuHaltPollTime(virDomainObj *dom,
return;
}
+
+static void
+qemuDomainGetStatsCpuLoadAvg(virDomainObj *dom,
+ virTypedParamList *params,
+ unsigned int privflags)
+{
+ qemuAgent *agent = NULL;
+ double load1m = 0;
+ double load5m = 0;
+ double load15m = 0;
+ int rc = 0;
+
+ if (!HAVE_JOB(privflags) ||
+ !virDomainObjIsActive(dom) ||
+ !qemuDomainAgentAvailable(dom, false))
+ return;
+
+ agent = qemuDomainObjEnterAgent(dom);
+ rc = qemuAgentGetLoadAvg(agent, &load1m, &load5m, &load15m, false);
+ qemuDomainObjExitAgent(dom, agent);
+
+ if (rc < 0)
+ return;
+
+ virTypedParamListAddDouble(params, load1m, "cpu.load.1m");
+ virTypedParamListAddDouble(params, load5m, "cpu.load.5m");
+ virTypedParamListAddDouble(params, load15m, "cpu.load.15m");
+}
+
+
static void
qemuDomainGetStatsCpu(virQEMUDriver *driver,
virDomainObj *dom,
@@ -16954,6 +16984,8 @@ qemuDomainGetStatsCpu(virQEMUDriver *driver,
qemuDomainGetStatsCpuCache(driver, dom, params);
qemuDomainGetStatsCpuHaltPollTime(dom, params, privflags);
+
+ qemuDomainGetStatsCpuLoadAvg(dom, params, privflags);
}
diff --git a/tests/qemuagenttest.c b/tests/qemuagenttest.c
index 328788024197..566571cf1107 100644
--- a/tests/qemuagenttest.c
+++ b/tests/qemuagenttest.c
@@ -1356,6 +1356,58 @@ testQemuAgentTimezone(const void *data)
virTypedParamsFree(params, nparams);
return ret;
}
+
+
+static const char testQemuAgentGetLoadAvgResponse[] =
+ "{"
+ " \"return\": {"
+ " \"load15m\": 0.03564453125,"
+ " \"load5m\": 0.064453125,"
+ " \"load1m\": 0.00390625"
+ " }"
+ "}";
+
+static int
+testQemuAgentGetLoadAvg(const void *data)
+{
+ virDomainXMLOption *xmlopt = (virDomainXMLOption *)data;
+ g_autoptr(qemuMonitorTest) test = qemuMonitorTestNewAgent(xmlopt);
+ double load1m = 0;
+ double load5m = 0;
+ double load15m = 0;
+
+ if (!test)
+ return -1;
+
+ if (qemuMonitorTestAddAgentSyncResponse(test) < 0)
+ return -1;
+
+ if (qemuMonitorTestAddItem(test, "guest-get-load",
+ testQemuAgentGetLoadAvgResponse) < 0)
+ return -1;
+
+ if (qemuAgentGetLoadAvg(qemuMonitorTestGetAgent(test),
+ &load1m, &load5m, &load15m, true) < 0)
+ return -1;
+
+#define VALIDATE_LOAD(value_, expected_) \
+ do { \
+ if (value_ != expected_) { \
+ virReportError(VIR_ERR_INTERNAL_ERROR, \
+ "Expected " #value_ " '%.11f', got '%.11f'", \
+ expected_, value_); \
+ return -1; \
+ } \
+ } while (0)
+
+ VALIDATE_LOAD(load1m, 0.00390625);
+ VALIDATE_LOAD(load5m, 0.064453125);
+ VALIDATE_LOAD(load15m, 0.03564453125);
+
+ return 0;
+}
+
+
static int
mymain(void)
{
@@ -1392,6 +1444,7 @@ mymain(void)
DO_TEST(Timezone);
DO_TEST(SSHKeys);
DO_TEST(GetDisks);
+ DO_TEST(GetLoadAvg);
DO_TEST(Timeout); /* Timeout should always be called last */
--
2.48.1
3 months, 2 weeks