Re: [PATCH v2 2/2] tests: migration-test: Add dirty ring test

[Peter Xu]

On Wed, Jun 16, 2021 at 02:55:55PM +0200, Juan Quintela wrote:
> Peter Xu <peterx@redhat.com> wrote:
> > Add dirty ring test if kernel supports it.  Add the dirty ring parameter on
> > source should be mostly enough, but let's change the dest too to make them
> > match always.
> >
> > Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
> > Signed-off-by: Peter Xu <peterx@redhat.com>
> 
> Reviewed-by: Juan Quintela <quintela@redhat.com>
> 
> Why we check with 4096 and not whatever the kernel returs?

Because the kernel returned value is the "maximum supported", while IMHO it's
better to test with some commonly used value (which should normally not the
maximum supported one; 4096 is suggested the default value to use so far in
e.g. qemu-options.hx).

> 
> So ....
> 
> > @@ -467,6 +471,8 @@ typedef struct {
> >      bool use_shmem;
> >      /* only launch the target process */
> >      bool only_target;
> > +    /* Use dirty ring if true; dirty logging otherwise */
> > +    bool use_dirty_ring;
> 
> make this an int
> 
> >      char *opts_source;
> >      char *opts_target;
> >  } MigrateStart;
> > @@ -573,11 +579,13 @@ static int test_migrate_start(QTestState **from, 
> > QTestState **to,
> >          shmem_opts = g_strdup("");
> >      }
> >  
> > -    cmd_source = g_strdup_printf("-accel kvm -accel tcg%s%s "
> > +    cmd_source = g_strdup_printf("-accel kvm%s -accel tcg%s%s "
> >                                   "-name source,debug-threads=on "
> >                                   "-m %s "
> >                                   "-serial file:%s/src_serial "
> >                                   "%s %s %s %s",
> > +                                 args->use_dirty_ring ?
> > +                                 ",dirty-ring-size=4096" : "",
> 
> check if it is > 0 here and otherwise pass it
> 
> >                                   machine_opts ? " -machine " : "",
> >                                   machine_opts ? machine_opts : "",
> >                                   memory_size, tmpfs,
> > @@ -587,12 +595,14 @@ static int test_migrate_start(QTestState **from, 
> > QTestState **to,
> >          *from = qtest_init(cmd_source);
> 
> > +static bool kvm_dirty_ring_supported(void)
> > +{
> > +#if defined(__linux__)
> > +    int ret, kvm_fd = open("/dev/kvm", O_RDONLY);
> > +
> > +    if (kvm_fd < 0) {
> > +        return false;
> > +    }
> > +
> > +    ret = ioctl(kvm_fd, KVM_CHECK_EXTENSION, KVM_CAP_DIRTY_LOG_RING);
> > +    close(kvm_fd);
> > +
> > +    /* We test with 4096 slots */
> > +    if (ret < 4096) {
> 
> And we return this value.
> 
> > +        return false;
> > +    }
> > +
> > +    return true;
> > +#else
> > +    return false;
> > +#endif
> > +}
> 
> Mental note, don't we have a function inside tests/ that check if
> a kvm feauture is there?
> 
> Quick check seems to imply no.
> 
> On the other hand, KVM_CHECK_EXTENSION is basically only used on
> kvm-all.c

Yeah, seems not there yet..  So we don't have guarantee to run this new test
because normally the host kernel is unpredictable, but it'll run on more hosts
as time flies and as people upgrade the kernels, meanwhile it can be run by
anyone who wants to regress dirty ring with a known new host kernel (so
e.g. for migration pull we can run migration-test with any host kernel newer
than 5.11+ and also run with root to cover all test cases there).

Thanks,

-- 
Peter Xu