[PATCH v2] tools: only fail validations if VIR_HOST_VALIDATE_FAIL is set
Michal Prívozník
mprivozn at redhat.com
Tue Jun 8 06:51:53 UTC 2021
On 6/7/21 6:22 PM, Fabiano Fidêncio wrote:
> Currently `virt-host-validate` will fail whenever one of its calls fail,
> regardless of virHostValidateLevel set.
>
> This behaviour is not optimal and makes it not exactly reliable as a
> command line tool as other tools or scripts using it would have to check
> its output to figure out whether something really failed or if a warning
> was mistakenly treated as failure.
>
> With this change, the behaviour of whether to fail or not, is defined by
> the caller of those functions, based on the virHostValidateLevel passed
> to them.
>
> https://gitlab.com/libvirt/libvirt/-/issues/175
>
> Signed-off-by: Fabiano Fidêncio <fabiano at fidencio.org>
> ---
> Changes since v1:
> * Replace the `goto out;` and the `out` labels by the
> `VIR_HOST_VALIDATE_FAILURE` macro
Yeah, this opened pandora's box...
> ---
> tools/virt-host-validate-common.c | 30 +++++++++++++++---------------
> tools/virt-host-validate-common.h | 14 ++++++++++++++
> 2 files changed, 29 insertions(+), 15 deletions(-)
>
> diff --git a/tools/virt-host-validate-common.c b/tools/virt-host-validate-common.c
> index 6dd851f07d..9412bb7514 100644
> --- a/tools/virt-host-validate-common.c
> +++ b/tools/virt-host-validate-common.c
> @@ -142,7 +142,7 @@ int virHostValidateDeviceExists(const char *hvname,
>
> if (access(dev_name, F_OK) < 0) {
> virHostMsgFail(level, "%s", hint);
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
>
> virHostMsgPass();
> @@ -159,7 +159,7 @@ int virHostValidateDeviceAccessible(const char *hvname,
>
> if (access(dev_name, R_OK|W_OK) < 0) {
> virHostMsgFail(level, "%s", hint);
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
>
> virHostMsgPass();
> @@ -180,7 +180,7 @@ int virHostValidateNamespace(const char *hvname,
>
> if (access(nspath, F_OK) < 0) {
> virHostMsgFail(level, "%s", hint);
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
>
> virHostMsgPass();
> @@ -264,17 +264,17 @@ int virHostValidateLinuxKernel(const char *hvname,
>
> if (STRNEQ(uts.sysname, "Linux")) {
> virHostMsgFail(level, "%s", hint);
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
>
> if (virParseVersionString(uts.release, &thisversion, true) < 0) {
> virHostMsgFail(level, "%s", hint);
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
>
> if (thisversion < version) {
> virHostMsgFail(level, "%s", hint);
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
Up until here the use of VIR_HOST_VALIDATE_FAILURE() is good.
> } else {
> virHostMsgPass();
> return 0;
> @@ -291,7 +291,7 @@ int virHostValidateCGroupControllers(const char *hvname,
> size_t i;
>
> if (virCgroupNew("/", -1, &group) < 0)
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
But this looks somewhat suspicious. What virCgroupNew() does is it
detects controllers and their mountpoints (and return this in @group).
Upon error (e.g. unable to open /proc/mounts or /proc/self/cgroups or
failure in their parsing) a libvirt error is reported. And this reminds
me of pandora's box. Firstly, the error object is never initialized
(i.e. there's no virInitialize() call from main(), or better said before
the first function that has potential of reporting an error). This leads
to virResetError() (called from virReportError()) to free() random
pointers.
Secondly, I'm unsure what to do in this case. I mean, caller told us
severity of this check (@level), so if we failed to initialize an
internal structure we should honour caller's wish and do/do not return
an error. BUT, with the way this code is currently written this function
exits early, not printing anything anywhere. E.g.:
QEMU: Checking if device /dev/net/tun exists : PASS
QEMU: Checking for device assignment IOMMU support : PASS
Whereas this is from a regular run:
QEMU: Checking if device /dev/net/tun exists : PASS
QEMU: Checking for cgroup 'cpu' controller support : PASS
QEMU: Checking for cgroup 'cpuacct' controller support : PASS
QEMU: Checking for cgroup 'cpuset' controller support : PASS
QEMU: Checking for cgroup 'memory' controller support : PASS
QEMU: Checking for cgroup 'devices' controller support : PASS
QEMU: Checking for cgroup 'blkio' controller support : PASS
QEMU: Checking for device assignment IOMMU support : PASS
This becomes more obvious [1]
>
> for (i = 0; i < VIR_CGROUP_CONTROLLER_LAST; i++) {
> int flag = 1 << i;
> @@ -303,7 +303,7 @@ int virHostValidateCGroupControllers(const char *hvname,
> virHostMsgCheck(hvname, "for cgroup '%s' controller support", cg_name);
>
> if (!virCgroupHasController(group, i)) {
> - ret = -1;
> + ret = VIR_HOST_VALIDATE_FAILURE(level);
> virHostMsgFail(level, "Enable '%s' in kernel Kconfig file or "
> "mount/enable cgroup controller in your system",
> cg_name);
> @@ -320,7 +320,7 @@ int virHostValidateCGroupControllers(const char *hvname G_GNUC_UNUSED,
> virHostValidateLevel level)
> {
> virHostMsgFail(level, "%s", "This platform does not support cgroups");
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
> #endif /* !__linux__ */
>
> @@ -354,7 +354,7 @@ int virHostValidateIOMMU(const char *hvname,
> "No ACPI DMAR table found, IOMMU either "
> "disabled in BIOS or not supported by this "
> "hardware platform");
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
> } else if (isAMD) {
> virHostMsgCheck(hvname, "%s", _("for device assignment IOMMU support"));
> @@ -366,7 +366,7 @@ int virHostValidateIOMMU(const char *hvname,
> "No ACPI IVRS table found, IOMMU either "
> "disabled in BIOS or not supported by this "
> "hardware platform");
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
> } else if (ARCH_IS_PPC64(arch)) {
> /* Empty Block */
> @@ -385,7 +385,7 @@ int virHostValidateIOMMU(const char *hvname,
> } else {
> virHostMsgFail(level,
> "Unknown if this platform has IOMMU support");
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
>
>
> @@ -404,7 +404,7 @@ int virHostValidateIOMMU(const char *hvname,
> "Add %s to kernel cmdline arguments", bootarg);
> else
> virHostMsgFail(level, "IOMMU capability not compiled into kernel.");
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
> }
> virHostMsgPass();
> return 0;
> @@ -468,7 +468,7 @@ int virHostValidateSecureGuests(const char *hvname,
> }
>
> if (virFileReadValueString(&cmdline, "/proc/cmdline") < 0)
> - return -1;
> + return VIR_HOST_VALIDATE_FAILURE(level);
1: here. IIUC, at the beginning of this function (not included in this
context) a message is printed out:
virHostMsgCheck(hvname, "%s", _("for secure guest support"));
but this 'return' makes it exit early (even without terminating the
line, corrupting the output).
So, long story short, I'm inclined to merge your patch and will post
some cleanups shortly.
Reviewed-by: Michal Privoznik <mprivozn at redhat.com>
Michal
More information about the libvir-list
mailing list