| Message ID | 1491986217-28215-1-git-send-email-weinan.z.li@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Apr 12, 2017 at 04:36:57PM +0800, Weinan Li wrote: > I915_GEM_GET_APERTURE ioctl is used to probe aperture size from userspace. > Some applications like OpenCL use this information to know how much GM > resource can it use. That's a userspace bug. > In gvt environment, each vm only use the ballooned > part of aperture, so we should return the actual aperture size exclude > the reserved part by balloon. > > I915_GEM_CONTEXT_GETPARAM ioctl query the I915_CONTEXT_PARAM_GTT_SIZE, we > also need to exclude the reserved part in GTT. > > Signed-off-by: Weinan Li <weinan.z.li@intel.com> > --- > drivers/gpu/drm/i915/i915_gem.c | 7 +++---- > drivers/gpu/drm/i915/i915_gem_context.c | 4 +++- > drivers/gpu/drm/i915/i915_vgpu.c | 18 +++++++++++++++++- > drivers/gpu/drm/i915/i915_vgpu.h | 2 ++ > 4 files changed, 25 insertions(+), 6 deletions(-) > > diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c > index 84ea249..b3fb424 100644 > --- a/drivers/gpu/drm/i915/i915_gem.c > +++ b/drivers/gpu/drm/i915/i915_gem.c > @@ -145,9 +145,8 @@ int i915_mutex_lock_interruptible(struct drm_device *dev) > struct i915_ggtt *ggtt = &dev_priv->ggtt; > struct drm_i915_gem_get_aperture *args = data; > struct i915_vma *vma; > - size_t pinned; > + size_t pinned = 0; > > - pinned = 0; > mutex_lock(&dev->struct_mutex); > list_for_each_entry(vma, &ggtt->base.active_list, vm_link) > if (i915_vma_is_pinned(vma)) > @@ -157,9 +156,9 @@ int i915_mutex_lock_interruptible(struct drm_device *dev) > pinned += vma->node.size; > mutex_unlock(&dev->struct_mutex); > > - args->aper_size = ggtt->base.total; > + args->aper_size = ggtt->base.total - > + intel_vgt_reserved_size_by_balloon(dev_priv); > args->aper_available_size = args->aper_size - pinned; > - > return 0; > } > > diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c > index 8bd0c49..9f3280d 100644 > --- a/drivers/gpu/drm/i915/i915_gem_context.c > +++ b/drivers/gpu/drm/i915/i915_gem_context.c > @@ -88,6 +88,7 @@ > #include <drm/drmP.h> > #include <drm/i915_drm.h> > #include "i915_drv.h" > +#include "i915_vgpu.h" > #include "i915_trace.h" > > #define ALL_L3_SLICES(dev) (1 << NUM_L3_SLICES(dev)) - 1 > @@ -1053,7 +1054,8 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data, > else if (to_i915(dev)->mm.aliasing_ppgtt) > args->value = to_i915(dev)->mm.aliasing_ppgtt->base.total; > else > - args->value = to_i915(dev)->ggtt.base.total; > + args->value = to_i915(dev)->ggtt.base.total - > + intel_vgt_reserved_size_by_balloon(dev->dev_private); > break; > case I915_CONTEXT_PARAM_NO_ERROR_CAPTURE: > args->value = i915_gem_context_no_error_capture(ctx); > diff --git a/drivers/gpu/drm/i915/i915_vgpu.c b/drivers/gpu/drm/i915/i915_vgpu.c > index 4ab8a97..ce722d8 100644 > --- a/drivers/gpu/drm/i915/i915_vgpu.c > +++ b/drivers/gpu/drm/i915/i915_vgpu.c > @@ -88,6 +88,11 @@ struct _balloon_info_ { > * graphic memory, 2/3 for unmappable graphic memory. > */ > struct drm_mm_node space[4]; > + /* > + * Total space size exclude ballooned named reserved_total, it's > + * invisible for vGPU. > + */ > + size_t reserved_total; What is size_t? > }; > > static struct _balloon_info_ bl_info; > @@ -116,6 +121,14 @@ void intel_vgt_deballoon(struct drm_i915_private *dev_priv) > memset(&bl_info, 0, sizeof(bl_info)); > } > > +size_t intel_vgt_reserved_size_by_balloon(struct drm_i915_private *dev_priv) > +{ > + if (!intel_vgpu_active(dev_priv)) > + return 0; > + > + return bl_info.reserved_total; > +} Or just return bl_info.reserved_total. Why is there a global here anyway? Better would be to track dev_priv->ggtt.reserved Then the core code becomes gtt_size = dev_priv->ggtt.total - dev_priv->ggtt.reserved; and doesn't need to know the identity of every possible consumer. -Chris
On ke, 2017-04-12 at 09:53 +0100, Chris Wilson wrote: > On Wed, Apr 12, 2017 at 04:36:57PM +0800, Weinan Li wrote: > > > > I915_GEM_GET_APERTURE ioctl is used to probe aperture size from userspace. > > Some applications like OpenCL use this information to know how much GM > > resource can it use. > > That's a userspace bug. Yes, a new property might be in place. I don't think we can go and change the meaning of a parameter just like that. <SNIP> > > @@ -116,6 +121,14 @@ void intel_vgt_deballoon(struct drm_i915_private *dev_priv) > > memset(&bl_info, 0, sizeof(bl_info)); > > } > > > > +size_t intel_vgt_reserved_size_by_balloon(struct drm_i915_private *dev_priv) > > +{ > > + if (!intel_vgpu_active(dev_priv)) > > + return 0; > > + > > + return bl_info.reserved_total; > > +} > > Or just return bl_info.reserved_total. > > Why is there a global here anyway? > > Better would be to track dev_priv->ggtt.reserved > > Then the core code becomes > gtt_size = dev_priv->ggtt.total - dev_priv->ggtt.reserved; > > and doesn't need to know the identity of every possible consumer. I was writing an e-mail about the same thing. So +1 on the idea. Regards, Joonas
> -----Original Message----- > From: Joonas Lahtinen [mailto:joonas.lahtinen@linux.intel.com] > Sent: Wednesday, April 12, 2017 6:19 PM > To: Chris Wilson <chris@chris-wilson.co.uk>; Li, Weinan Z > <weinan.z.li@intel.com> > Cc: intel-gfx@lists.freedesktop.org; intel-gvt-dev@lists.freedesktop.org > Subject: Re: [Intel-gfx] [PATCH] drm/i915/gvt: return the actual aperture size > under gvt environment > > On ke, 2017-04-12 at 09:53 +0100, Chris Wilson wrote: > > On Wed, Apr 12, 2017 at 04:36:57PM +0800, Weinan Li wrote: > > > > > > I915_GEM_GET_APERTURE ioctl is used to probe aperture size from > userspace. > > > Some applications like OpenCL use this information to know how much > > > GM resource can it use. > > > > That's a userspace bug. > > Yes, a new property might be in place. I don't think we can go and change the > meaning of a parameter just like that. > > <SNIP> > Here I don’t want to change the meaning of I915_GEM_GET_APERTURE, but for the ioctl, We need to return the actual available aperture size exclude the reserved space by GVT balloon. > > > @@ -116,6 +121,14 @@ void intel_vgt_deballoon(struct drm_i915_private > *dev_priv) > > > memset(&bl_info, 0, sizeof(bl_info)); > > > } > > > > > > +size_t intel_vgt_reserved_size_by_balloon(struct drm_i915_private > > > +*dev_priv) { > > > + if (!intel_vgpu_active(dev_priv)) > > > + return 0; > > > + > > > + return bl_info.reserved_total; > > > +} > > > > Or just return bl_info.reserved_total. > > > > Why is there a global here anyway? > > > > Better would be to track dev_priv->ggtt.reserved > > > > Then the core code becomes > > gtt_size = dev_priv->ggtt.total - dev_priv->ggtt.reserved; > > > > and doesn't need to know the identity of every possible consumer. > > I was writing an e-mail about the same thing. So +1 on the idea. > Agreed. > Regards, Joonas > -- > Joonas Lahtinen > Open Source Technology Center > Intel Corporation
On to, 2017-04-13 at 01:01 +0000, Li, Weinan Z wrote: > > > > -----Original Message----- > > From: Joonas Lahtinen [mailto:joonas.lahtinen@linux.intel.com] > > Sent: Wednesday, April 12, 2017 6:19 PM > > To: Chris Wilson <chris@chris-wilson.co.uk>; Li, Weinan Z > > <weinan.z.li@intel.com> > > Cc: intel-gfx@lists.freedesktop.org; intel-gvt-dev@lists.freedesktop.org > > Subject: Re: [Intel-gfx] [PATCH] drm/i915/gvt: return the actual aperture size > > under gvt environment > > > > On ke, 2017-04-12 at 09:53 +0100, Chris Wilson wrote: > > > > > > On Wed, Apr 12, 2017 at 04:36:57PM +0800, Weinan Li wrote: > > > > > > > > > > > > I915_GEM_GET_APERTURE ioctl is used to probe aperture size from > > userspace. > > > > > > > > > > > Some applications like OpenCL use this information to know how much > > > > GM resource can it use. > > > > > > That's a userspace bug. > > > > Yes, a new property might be in place. I don't think we can go and change the > > meaning of a parameter just like that. > > > > <SNIP> > > > Here I don’t want to change the meaning of I915_GEM_GET_APERTURE, but for the ioctl, > We need to return the actual available aperture size exclude the reserved space by GVT balloon. IOCTLs represent the ABI contract we have with userspace. It has previously returned size of the aperture, so we can't change it to be something else (like the usable size of aperture as proposed here). Somebody might be doing an assert that any address in aperture is below I915_GEM_GET_APERTURE returned value, which has previously been correct, but would be broken after this change. There are also potentially other things consuming the aperture than VGT ballooning, so the UMDs would still be misbehaving. Shouldn't they rather be doing these decisions based on aper_available_size? Regards, Joonas
> -----Original Message----- > From: Joonas Lahtinen [mailto:joonas.lahtinen@linux.intel.com] > Sent: Thursday, April 13, 2017 6:11 PM > To: Li, Weinan Z <weinan.z.li@intel.com>; Chris Wilson <chris@chris- > wilson.co.uk> > Cc: intel-gfx@lists.freedesktop.org; intel-gvt-dev@lists.freedesktop.org > Subject: Re: [Intel-gfx] [PATCH] drm/i915/gvt: return the actual aperture size > under gvt environment > > On to, 2017-04-13 at 01:01 +0000, Li, Weinan Z wrote: > > > > > > -----Original Message----- > > > From: Joonas Lahtinen [mailto:joonas.lahtinen@linux.intel.com] > > > Sent: Wednesday, April 12, 2017 6:19 PM > > > To: Chris Wilson <chris@chris-wilson.co.uk>; Li, Weinan Z > > > <weinan.z.li@intel.com> > > > Cc: intel-gfx@lists.freedesktop.org; > > > intel-gvt-dev@lists.freedesktop.org > > > Subject: Re: [Intel-gfx] [PATCH] drm/i915/gvt: return the actual > > > aperture size under gvt environment > > > > > > On ke, 2017-04-12 at 09:53 +0100, Chris Wilson wrote: > > > > > > > > On Wed, Apr 12, 2017 at 04:36:57PM +0800, Weinan Li wrote: > > > > > > > > > > > > > > > I915_GEM_GET_APERTURE ioctl is used to probe aperture size from > > > userspace. > > > > > > > > > > > > > > Some applications like OpenCL use this information to know how > > > > > much GM resource can it use. > > > > > > > > That's a userspace bug. > > > > > > Yes, a new property might be in place. I don't think we can go and > > > change the meaning of a parameter just like that. > > > > > > <SNIP> > > > > > Here I don’t want to change the meaning of I915_GEM_GET_APERTURE, but > > for the ioctl, We need to return the actual available aperture size exclude the > reserved space by GVT balloon. > > IOCTLs represent the ABI contract we have with userspace. It has previously > returned size of the aperture, so we can't change it to be something else (like > the usable size of aperture as proposed here). > > Somebody might be doing an assert that any address in aperture is below > I915_GEM_GET_APERTURE returned value, which has previously been correct, > but would be broken after this change. There are also potentially other things > consuming the aperture than VGT ballooning, so the UMDs would still be > misbehaving. > > Shouldn't they rather be doing these decisions based on aper_available_size? > Known your mean, if we return the value as below: - args->aper_size = ggtt->base.total; + args->aper_size = ggtt->base.total - ggtt->base.reserved; Then userspace may use 'args->aper_size' as the MAX aperture addr, it may cause other issues. In GVT with balloon the aperture addr still be from 0 to ggtt->base.total. If it's expected behavior, change the available aperture size may avoid this. args->aper_size = ggtt->base.total; args->aper_available_size = args->aper_size - ggtt->base.reserved - pinned; > Regards, Joonas > -- > Joonas Lahtinen > Open Source Technology Center > Intel Corporation
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 84ea249..b3fb424 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -145,9 +145,8 @@ int i915_mutex_lock_interruptible(struct drm_device *dev) struct i915_ggtt *ggtt = &dev_priv->ggtt; struct drm_i915_gem_get_aperture *args = data; struct i915_vma *vma; - size_t pinned; + size_t pinned = 0; - pinned = 0; mutex_lock(&dev->struct_mutex); list_for_each_entry(vma, &ggtt->base.active_list, vm_link) if (i915_vma_is_pinned(vma)) @@ -157,9 +156,9 @@ int i915_mutex_lock_interruptible(struct drm_device *dev) pinned += vma->node.size; mutex_unlock(&dev->struct_mutex); - args->aper_size = ggtt->base.total; + args->aper_size = ggtt->base.total - + intel_vgt_reserved_size_by_balloon(dev_priv); args->aper_available_size = args->aper_size - pinned; - return 0; } diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index 8bd0c49..9f3280d 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -88,6 +88,7 @@ #include <drm/drmP.h> #include <drm/i915_drm.h> #include "i915_drv.h" +#include "i915_vgpu.h" #include "i915_trace.h" #define ALL_L3_SLICES(dev) (1 << NUM_L3_SLICES(dev)) - 1 @@ -1053,7 +1054,8 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data, else if (to_i915(dev)->mm.aliasing_ppgtt) args->value = to_i915(dev)->mm.aliasing_ppgtt->base.total; else - args->value = to_i915(dev)->ggtt.base.total; + args->value = to_i915(dev)->ggtt.base.total - + intel_vgt_reserved_size_by_balloon(dev->dev_private); break; case I915_CONTEXT_PARAM_NO_ERROR_CAPTURE: args->value = i915_gem_context_no_error_capture(ctx); diff --git a/drivers/gpu/drm/i915/i915_vgpu.c b/drivers/gpu/drm/i915/i915_vgpu.c index 4ab8a97..ce722d8 100644 --- a/drivers/gpu/drm/i915/i915_vgpu.c +++ b/drivers/gpu/drm/i915/i915_vgpu.c @@ -88,6 +88,11 @@ struct _balloon_info_ { * graphic memory, 2/3 for unmappable graphic memory. */ struct drm_mm_node space[4]; + /* + * Total space size exclude ballooned named reserved_total, it's + * invisible for vGPU. + */ + size_t reserved_total; }; static struct _balloon_info_ bl_info; @@ -116,6 +121,14 @@ void intel_vgt_deballoon(struct drm_i915_private *dev_priv) memset(&bl_info, 0, sizeof(bl_info)); } +size_t intel_vgt_reserved_size_by_balloon(struct drm_i915_private *dev_priv) +{ + if (!intel_vgpu_active(dev_priv)) + return 0; + + return bl_info.reserved_total; +} + static int vgt_balloon_space(struct i915_ggtt *ggtt, struct drm_mm_node *node, unsigned long start, unsigned long end) @@ -183,7 +196,7 @@ int intel_vgt_balloon(struct drm_i915_private *dev_priv) unsigned long mappable_base, mappable_size, mappable_end; unsigned long unmappable_base, unmappable_size, unmappable_end; - int ret; + int ret, i; if (!intel_vgpu_active(dev_priv)) return 0; @@ -242,6 +255,9 @@ int intel_vgt_balloon(struct drm_i915_private *dev_priv) goto err; } + for (i = 0; i < ARRAY_SIZE(bl_info.space); i++) + bl_info.reserved_total += bl_info.space[i].size; + DRM_INFO("VGT balloon successfully\n"); return 0; diff --git a/drivers/gpu/drm/i915/i915_vgpu.h b/drivers/gpu/drm/i915/i915_vgpu.h index 3c3b2d2..e776580 100644 --- a/drivers/gpu/drm/i915/i915_vgpu.h +++ b/drivers/gpu/drm/i915/i915_vgpu.h @@ -29,5 +29,7 @@ void i915_check_vgpu(struct drm_i915_private *dev_priv); int intel_vgt_balloon(struct drm_i915_private *dev_priv); void intel_vgt_deballoon(struct drm_i915_private *dev_priv); +size_t intel_vgt_reserved_size_by_balloon( + struct drm_i915_private *dev_priv); #endif /* _I915_VGPU_H_ */
I915_GEM_GET_APERTURE ioctl is used to probe aperture size from userspace. Some applications like OpenCL use this information to know how much GM resource can it use. In gvt environment, each vm only use the ballooned part of aperture, so we should return the actual aperture size exclude the reserved part by balloon. I915_GEM_CONTEXT_GETPARAM ioctl query the I915_CONTEXT_PARAM_GTT_SIZE, we also need to exclude the reserved part in GTT. Signed-off-by: Weinan Li <weinan.z.li@intel.com> --- drivers/gpu/drm/i915/i915_gem.c | 7 +++---- drivers/gpu/drm/i915/i915_gem_context.c | 4 +++- drivers/gpu/drm/i915/i915_vgpu.c | 18 +++++++++++++++++- drivers/gpu/drm/i915/i915_vgpu.h | 2 ++ 4 files changed, 25 insertions(+), 6 deletions(-)