[PATCH] drm/i915/gvt: Introduce vGPU mm notifier

[Tian, Kevin]

Then using a RFC prefix is clearer.

> -----Original Message-----
> From: intel-gvt-dev [mailto:intel-gvt-dev-bounces at lists.freedesktop.org] On
> Behalf Of Zhi Wang
> Sent: Monday, April 24, 2017 1:21 PM
> To: Zhenyu Wang <zhenyuw at linux.intel.com>
> Cc: intel-gvt-dev at lists.freedesktop.org
> Subject: Re: [PATCH] drm/i915/gvt: Introduce vGPU mm notifier
> 
> Sure!, I'm collecting the comments at this time. :)
> 
> 于 04/24/17 10:50, Zhenyu Wang 写道:
> > On 2017.04.21 23:32:28 +0800, Zhi Wang wrote:
> >> This patch introduces vGPU mm notifier for clients who may wish to track
> >> guest changing its graphics memory mapping in a vGPU GGTT/PPGTT mm.
> >>
> >> Signed-off-by: Zhi Wang <zhi.a.wang at intel.com>
> >> ---
> >
> > We might wait to merge this until real user will be ready.
> >
> >> +
> >> +/**
> >> + * intel_vgpu_activate_mm_notifier - activate a vGPU mm notifier
> >> + * @n: a vGPU mm notifier
> >> + *
> >> + * This function is used to start tracking the GMADR mapping change.
> >> + *
> >> + * Returns:
> >> + * Zero on success, negative error code in pointer if failed.
> >> + */
> >> +int intel_vgpu_activate_mm_notifier(struct intel_vgpu_mm_notifier *n)
> >> +{
> >> +	if (n->active)
> >> +		return -EEXIST;
> >> +
> >> +	n->active = true;
> >> +	interval_tree_insert(&n->node, &n->mm->mm_notifiers);
> >> +
> >> +	return 0;
> >> +}
> >
> > I think you should add start/last setting here for this notifier
> > instead of leaving out to caller, which will be consistent then.
> >
> >> +
> >> +/**
> >> + * intel_vgpu_deactivate_mm_notifier - de-activate a vGPU mm notifier
> >> + * @n: a vGPU mm notifier
> >> + *
> >> + * This function is used to stop tracking the GMADR mapping change. If a
> >> + * client wish to modify the tracking range, it has to stop the tracking
> >> + * first and then start the tracking again after modifying the range.
> >> + *
> >> + */
> >> +void intel_vgpu_deactivate_notifier(struct intel_vgpu_mm_notifier *n)
> >> +{
> >> +	interval_tree_remove(&n->node, &n->mm->mm_notifiers);
> >> +	n->active = false;
> >> +}
> >> diff --git a/drivers/gpu/drm/i915/gvt/gtt.h
> b/drivers/gpu/drm/i915/gvt/gtt.h
> >> index f88eb5e..95cfe7b 100644
> >> --- a/drivers/gpu/drm/i915/gvt/gtt.h
> >> +++ b/drivers/gpu/drm/i915/gvt/gtt.h
> >> @@ -121,6 +121,18 @@ typedef enum {
> >>   	GTT_TYPE_MAX,
> >>   } intel_gvt_gtt_type_t;
> >>
> >> +struct intel_vgpu_mm_notifier {
> >> +	bool active;
> >> +
> >> +	struct intel_vgpu_mm *mm;
> >> +	int (*mapping_change)(struct intel_vgpu_mm_notifier *, u64, u64);
> >> +	struct interval_tree_node node;
> >> +};
> >> +
> >> +extern int intel_vgpu_activate_mm_notifier(struct
> intel_vgpu_mm_notifier *n);
> >> +
> >> +extern void intel_vgpu_deactivate_mm_notifier(struct
> intel_vgpu_mm_notifier *n);
> >> +
> >>   struct intel_vgpu_mm {
> >>   	int type;
> >>   	bool initialized;
> >> @@ -141,6 +153,7 @@ struct intel_vgpu_mm {
> >>   	atomic_t pincount;
> >>   	struct list_head lru_list;
> >>   	struct intel_vgpu *vgpu;
> >> +	struct rb_root mm_notifiers;
> >>   };
> >
> > You also need to handle notifier tree cleanup when vgpu mm destroy.
> >
> >>
> >>   extern struct intel_gvt_gtt_entry *intel_vgpu_mm_get_entry(
> >> diff --git a/drivers/gpu/drm/i915/gvt/gvt.h
> b/drivers/gpu/drm/i915/gvt/gvt.h
> >> index becae2f..b3629404 100644
> >> --- a/drivers/gpu/drm/i915/gvt/gvt.h
> >> +++ b/drivers/gpu/drm/i915/gvt/gvt.h
> >> @@ -33,6 +33,8 @@
> >>   #ifndef _GVT_H_
> >>   #define _GVT_H_
> >>
> >> +#include <linux/interval_tree.h>
> >> +
> >>   #include "debug.h"
> >>   #include "hypercall.h"
> >>   #include "mmio.h"
> >> --
> >> 2.7.4
> >>
> >> _______________________________________________
> >> intel-gvt-dev mailing list
> >> intel-gvt-dev at lists.freedesktop.org
> >> https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev
> >
> _______________________________________________
> intel-gvt-dev mailing list
> intel-gvt-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev