Re: [PATCH v2] qemu: Introduce caching whether /dev/kvm is accessible

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Oct 30, 2018 at 09:13:50AM +0100, Michal Privoznik wrote:
> On 10/29/2018 06:34 PM, Marc Hartmayer wrote:
> > Introduce caching whether /dev/kvm is usable as the QEMU user:QEMU
> > group. This reduces the overhead of the QEMU capabilities cache
> > lookup. Before this patch there were many fork() calls used for
> > checking whether /dev/kvm is accessible. Now we store the result
> > whether /dev/kvm is accessible or not and we only need to re-run the
> > virFileAccessibleAs check if the ctime of /dev/kvm has changed.
> > 
> > Suggested-by: Daniel P. Berrangé <berrange@xxxxxxxxxx>
> > Signed-off-by: Marc Hartmayer <mhartmay@xxxxxxxxxxxxx>
> > ---
> >  src/qemu/qemu_capabilities.c | 54 ++++++++++++++++++++++++++++++++++--
> >  1 file changed, 52 insertions(+), 2 deletions(-)
> > 
> > diff --git a/src/qemu/qemu_capabilities.c b/src/qemu/qemu_capabilities.c
> > index e228f52ec0bb..85516954149b 100644
> > --- a/src/qemu/qemu_capabilities.c
> > +++ b/src/qemu/qemu_capabilities.c
> > @@ -3238,6 +3238,10 @@ struct _virQEMUCapsCachePriv {
> >      virArch hostArch;
> >      unsigned int microcodeVersion;
> >      char *kernelVersion;
> > +
> > +    /* cache whether /dev/kvm is usable as runUid:runGuid */
> > +    virTristateBool kvmUsable;
> > +    time_t kvmCtime;
> >  };
> >  typedef struct _virQEMUCapsCachePriv virQEMUCapsCachePriv;
> >  typedef virQEMUCapsCachePriv *virQEMUCapsCachePrivPtr;
> > @@ -3824,6 +3828,52 @@ virQEMUCapsSaveFile(void *data,
> >  }
> >  
> >  
> > +/* Determine whether '/dev/kvm' is usable as QEMU user:QEMU group. */
> > +static bool
> > +virQEMUCapsKVMUsable(virQEMUCapsCachePrivPtr priv)
> > +{
> > +    struct stat sb;
> > +    static const char *kvm_device = "/dev/kvm";
> > +    virTristateBool value;
> > +    virTristateBool cached_value = priv->kvmUsable;
> > +    time_t kvm_ctime;
> > +    time_t cached_kvm_ctime = priv->kvmCtime;
> > +
> > +    if (stat(kvm_device, &sb) < 0) {
> > +        virReportSystemError(errno,
> > +                             _("Failed to stat %s"), kvm_device);
> > +        return false;
> > +    }
> > +    kvm_ctime = sb.st_ctime;
> 
> This doesn't feel right. /dev/kvm ctime is changed every time qemu is
> started or powered off (try running stat over it before and after a
> domain is started/shut off). So effectively we will fork more often than
> we would think. Should we cache inode number instead? Because for all
> that we care is simply if the file is there.

Urgh, that is a bit strange and not the usual semantics for timestamps :-(

We can't stat the inode - the code was explicitly trying to cope with the
way /dev/kvm can change permissions when udev rules get applied. We would
have to compare the user, group, permissions mask and even ACL, or a hash
of those.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|

--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list




[Index of Archives]     [Virt Tools]     [Libvirt Users]     [Lib OS Info]     [Fedora Users]     [Fedora Desktop]     [Fedora SELinux]     [Big List of Linux Books]     [Yosemite News]     [KDE Users]     [Fedora Tools]

  Powered by Linux