[Intel-gfx] [PATCH i-g-t v2] igt/drv_module_reload: Revamp fault-injection

[Chris Wilson]

Quoting Imre Deak (2018-06-06 19:00:52)
> On Wed, Jun 06, 2018 at 06:42:14PM +0100, Chris Wilson wrote:
> > The current method of checking for a failed module load is flawed, as we
> > only report the error on probing it is not being reported back by
> > modprobe. So we have to dig inside the module_parameters while the
> > module is still loaded to discover the error.
> > 
> > v2: Expect i915.inject_load_failure to be zero on success
> > 
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > Cc: Michał Winiarski <michal.winiarski at intel.com>
> > Cc: Imre Deak <imre.deak at intel.com>
> > Reviewed-by: Michał Winiarski <michal.winiarski at intel.com>
> > ---
> >  tests/drv_module_reload.c | 45 ++++++++++++++++++++++++++++++++++-----
> >  1 file changed, 40 insertions(+), 5 deletions(-)
> > 
> > diff --git a/tests/drv_module_reload.c b/tests/drv_module_reload.c
> > index 092982960..e18aaea5e 100644
> > --- a/tests/drv_module_reload.c
> > +++ b/tests/drv_module_reload.c
> > @@ -234,6 +234,38 @@ reload(const char *opts_i915)
> >       return err;
> >  }
> >  
> > +static int open_parameters(const char *module_name)
> > +{
> > +     char path[256];
> > +
> > +     snprintf(path, sizeof(path), "/sys/module/%s/parameters", module_name);
> > +     return open(path, O_RDONLY);
> > +}
> > +
> > +static int
> > +inject_fault(const char *module_name, const char *opt, int fault)
> > +{
> > +     char buf[1024];
> > +     int dir;
> > +
> > +     igt_assert(fault > 0);
> > +     snprintf(buf, sizeof(buf), "%s=%d", opt, fault);
> > +
> > +     if (igt_kmod_load(module_name, buf)) {
> > +             igt_warn("Failed to load module '%s' with options '%s'\n",
> > +                      module_name, buf);
> > +             return 1;
> > +     }
> > +
> > +     dir = open_parameters(module_name);
> > +     igt_sysfs_scanf(dir, opt, "%d", &fault);
> 
> Just a note for later: in case we switch to async probing we'll have to
> wait somehow until the probe completed. One way would be to disable
> async probing for this test, not sure if that could hide some problem
> though.

modprobe already waits for async probes (via async_synchronize_full()
before it returns to userspace). Afaict, that async flag only has any
relevance for builtins. If you want to parallel module loading, you need
to fork.
-Chris