[Intel-gfx] [PATCH 2/2] drm/i915: Add ring_mask module parameter

[Summers, Stuart]

On Wed, 2019-08-28 at 10:34 +0100, Chris Wilson wrote:
> Quoting Stuart Summers (2019-08-28 00:14:35)
> > Add a new module parameter, ring_mask, to allow for disabling
> > engines during i915 load. This mask follows the intel_engine_id
> > enum and can be used to hide specified engines from i915 and
> > from userspace.
> > 
> > Signed-off-by: Stuart Summers <stuart.summers at intel.com>
> > ---
> >  drivers/gpu/drm/i915/i915_params.c       | 3 +++
> >  drivers/gpu/drm/i915/i915_params.h       | 3 ++-
> >  drivers/gpu/drm/i915/intel_device_info.c | 6 ++++++
> >  3 files changed, 11 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/i915_params.c
> > b/drivers/gpu/drm/i915/i915_params.c
> > index 296452f9efe4..6a17fe6ea3a2 100644
> > --- a/drivers/gpu/drm/i915/i915_params.c
> > +++ b/drivers/gpu/drm/i915/i915_params.c
> > @@ -178,6 +178,9 @@ i915_param_named(enable_gvt, bool, 0400,
> >         "Enable support for Intel GVT-g graphics virtualization
> > host support(default:false)");
> >  #endif
> >  
> > +i915_param_named_unsafe(ring_mask, uint, 0400,
> > +       "Mask of engine rings to enable. (default: all supported
> > engines enabled)");
> > +
> >  static __always_inline void _print_param(struct drm_printer *p,
> >                                          const char *name,
> >                                          const char *type,
> > diff --git a/drivers/gpu/drm/i915/i915_params.h
> > b/drivers/gpu/drm/i915/i915_params.h
> > index d29ade3b7de6..d803bf5faac4 100644
> > --- a/drivers/gpu/drm/i915/i915_params.h
> > +++ b/drivers/gpu/drm/i915/i915_params.h
> > @@ -77,7 +77,8 @@ struct drm_printer;
> >         param(bool, verbose_state_checks, true) \
> >         param(bool, nuclear_pageflip, false) \
> >         param(bool, enable_dp_mst, true) \
> > -       param(bool, enable_gvt, false)
> > +       param(bool, enable_gvt, false) \
> > +       param(unsigned int, ring_mask, (unsigned int)-1)
> >  
> >  #define MEMBER(T, member, ...) T member;
> >  struct i915_params {
> > diff --git a/drivers/gpu/drm/i915/intel_device_info.c
> > b/drivers/gpu/drm/i915/intel_device_info.c
> > index caef01b1da23..f833280f2ef6 100644
> > --- a/drivers/gpu/drm/i915/intel_device_info.c
> > +++ b/drivers/gpu/drm/i915/intel_device_info.c
> > @@ -993,10 +993,16 @@ void intel_device_info_init_mmio(struct
> > drm_i915_private *dev_priv)
> >         u32 media_fuse;
> >         u16 vdbox_mask;
> >         u16 vebox_mask;
> > +       u32 orig_engine_mask = info->engine_mask;
> >  
> >         if (INTEL_GEN(dev_priv) < 11)
> >                 return;
> >  
> > +       info->engine_mask &= i915_modparams.ring_mask;
> 
> The challenge is we can't use our engine_mask for this as that is an
> internal bitfield that may change overtime -- and we definitely don't
> want the modparam to disable different engines on different kernel
> releases.

Yeah, it's a valid concern. I was thinking we might just put this
expectation on the user/debugger given this should really only be used
under duress. I'll rethink it and see if I can come up with something a
little safer.

> 
> > +       if (info->engine_mask != orig_engine_mask)
> > +               DRM_WARN("loading with reduced engine mask 0x%x\n",
> > +                        info->engine_mask);
> 
> It's operating under direct control from the user, so dev_notice (as
> we
> treat it as a quirk).

Makes sense. I'll make the change in the next revision.

Thanks,
Stuart

> -Chris
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3270 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20190828/b2a5809b/attachment-0001.bin>