RE: [PATCH v6 2/6] drm/i915/gvt: OpRegion support for GVT-g

[Chen, Xiaoguang]

Hi 

>-----Original Message-----
>From: intel-gvt-dev [mailto:intel-gvt-dev-bounces@xxxxxxxxxxxxxxxxxxxxx] On
>Behalf Of Zhenyu Wang
>Sent: Wednesday, May 31, 2017 12:47 PM
>To: Chen, Xiaoguang <xiaoguang.chen@xxxxxxxxx>
>Cc: Tian, Kevin <kevin.tian@xxxxxxxxx>; intel-gfx@xxxxxxxxxxxxxxxxxxxxx; linux-
>kernel@xxxxxxxxxxxxxxx; zhenyuw@xxxxxxxxxxxxxxx; chris@xxxxxxxxxxxxxxxxxx;
>alex.williamson@xxxxxxxxxx; kraxel@xxxxxxxxxx; Niu, Bing
><bing.niu@xxxxxxxxx>; intel-gvt-dev@xxxxxxxxxxxxxxxxxxxxx; Wang, Zhi A
><zhi.a.wang@xxxxxxxxx>; Lv, Zhiyuan <zhiyuan.lv@xxxxxxxxx>
>Subject: Re: [PATCH v6 2/6] drm/i915/gvt: OpRegion support for GVT-g
>
>On 2017.05.27 16:38:48 +0800, Xiaoguang Chen wrote:
>> OpRegion is needed to support display related operation for intel
>> vgpu.
>>
>> A vfio device region is added to intel vgpu to deliver the host
>> OpRegion information to user space so user space can construct the
>> OpRegion for vgpu.
>>
>> Signed-off-by: Bing Niu <bing.niu@xxxxxxxxx>
>> Signed-off-by: Xiaoguang Chen <xiaoguang.chen@xxxxxxxxx>
>> ---
>>  drivers/gpu/drm/i915/gvt/kvmgt.c    | 97
>+++++++++++++++++++++++++++++++++++++
>>  drivers/gpu/drm/i915/gvt/opregion.c |  8 ++-
>>  2 files changed, 104 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c
>> b/drivers/gpu/drm/i915/gvt/kvmgt.c
>> index 3c6a02b..389f072 100644
>> --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
>> +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
>> @@ -53,6 +53,8 @@ static const struct intel_gvt_ops *intel_gvt_ops;
>> #define VFIO_PCI_INDEX_TO_OFFSET(index) ((u64)(index) <<
>VFIO_PCI_OFFSET_SHIFT)
>>  #define VFIO_PCI_OFFSET_MASK    (((u64)(1) << VFIO_PCI_OFFSET_SHIFT) - 1)
>>
>> +#define OPREGION_SIGNATURE "IntelGraphicsMem"
>> +
>>  struct vfio_region;
>>  struct intel_vgpu_regops {
>>  	size_t (*rw)(struct intel_vgpu *vgpu, char *buf, @@ -436,6 +438,92
>> @@ static void kvmgt_protect_table_del(struct kvmgt_guest_info *info,
>>  	}
>>  }
>>
>> +static size_t intel_vgpu_reg_rw_opregion(struct intel_vgpu *vgpu, char *buf,
>> +		size_t count, loff_t *ppos, bool iswrite) {
>> +	unsigned int i = VFIO_PCI_OFFSET_TO_INDEX(*ppos) -
>> +			VFIO_PCI_NUM_REGIONS;
>> +	void *base = vgpu->vdev.region[i].data;
>> +	loff_t pos = *ppos & VFIO_PCI_OFFSET_MASK;
>> +
>> +	if (pos >= vgpu->vdev.region[i].size || iswrite) {
>> +		gvt_vgpu_err("invalid op or offset for Intel vgpu OpRegion\n");
>> +		return -EINVAL;
>> +	}
>> +	count = min(count, (size_t)(vgpu->vdev.region[i].size - pos));
>> +	memcpy(buf, base + pos, count);
>> +
>> +	return count;
>> +}
>> +
>> +static void intel_vgpu_reg_release_opregion(struct intel_vgpu *vgpu,
>> +		struct vfio_region *region)
>> +{
>> +	memunmap(region->data);
>> +}
>> +
>> +static const struct intel_vgpu_regops intel_vgpu_regops_opregion = {
>> +	.rw = intel_vgpu_reg_rw_opregion,
>> +	.release = intel_vgpu_reg_release_opregion, };
>> +
>> +static int intel_vgpu_register_reg(struct intel_vgpu *vgpu,
>> +		unsigned int type, unsigned int subtype,
>> +		const struct intel_vgpu_regops *ops,
>> +		size_t size, u32 flags, void *data) {
>> +	struct vfio_region *region;
>> +
>> +	region = krealloc(vgpu->vdev.region,
>> +			(vgpu->vdev.num_regions + 1) * sizeof(*region),
>> +			GFP_KERNEL);
>> +	if (!region)
>> +		return -ENOMEM;
>> +
>> +	vgpu->vdev.region = region;
>> +	vgpu->vdev.region[vgpu->vdev.num_regions].type = type;
>> +	vgpu->vdev.region[vgpu->vdev.num_regions].subtype = subtype;
>> +	vgpu->vdev.region[vgpu->vdev.num_regions].ops = ops;
>> +	vgpu->vdev.region[vgpu->vdev.num_regions].size = size;
>> +	vgpu->vdev.region[vgpu->vdev.num_regions].flags = flags;
>> +	vgpu->vdev.region[vgpu->vdev.num_regions].data = data;
>> +	vgpu->vdev.num_regions++;
>> +
>> +	return 0;
>> +}
>> +
>> +static int intel_vgpu_reg_init_opregion(struct intel_vgpu *vgpu) {
>> +	unsigned int addr;
>> +	void *base;
>> +	int ret;
>> +
>> +	addr = vgpu->gvt->opregion.opregion_pa;
>> +	if (!addr || !(~addr))
>> +		return -ENODEV;
>> +
>> +	base = memremap(addr, OPREGION_SIZE, MEMREMAP_WB);
>> +	if (!base)
>> +		return -ENOMEM;
>> +
>> +	if (memcmp(base, OPREGION_SIGNATURE, 16)) {
>> +		memunmap(base);
>> +		return -EINVAL;
>> +	}
>> +
>> +	ret = intel_vgpu_register_reg(vgpu,
>> +			PCI_VENDOR_ID_INTEL |
>VFIO_REGION_TYPE_PCI_VENDOR_TYPE,
>> +			VFIO_REGION_SUBTYPE_INTEL_IGD_OPREGION,
>> +			&intel_vgpu_regops_opregion, OPREGION_SIZE,
>> +			VFIO_REGION_INFO_FLAG_READ, base);
>> +	if (ret) {
>> +		memunmap(base);
>> +		return ret;
>> +	}
>> +
>> +	return ret;
>> +}
>> +
>>  static int intel_vgpu_create(struct kobject *kobj, struct mdev_device
>> *mdev)  {
>>  	struct intel_vgpu *vgpu = NULL;
>> @@ -467,6 +555,15 @@ static int intel_vgpu_create(struct kobject *kobj,
>struct mdev_device *mdev)
>>  	vgpu->vdev.mdev = mdev;
>>  	mdev_set_drvdata(mdev, vgpu);
>>
>> +	ret = intel_vgpu_reg_init_opregion(vgpu);
>> +	if (ret) {
>> +		gvt_vgpu_err("create OpRegion failed\n");
>> +		goto out;
>> +	}
>
>Still need to handle error path for created vgpu.
Just checked the code, if initialize the opregion failed we should first release vfio/mdev releated work(maybe call intel_vgpu_release function)  and then destroy the vgpu. Will update in the next version.

>
>> +
>> +	gvt_dbg_core("create OpRegion succeeded for mdev:%s\n",
>> +			dev_name(mdev_dev(mdev)));
>> +
>>  	gvt_dbg_core("intel_vgpu_create succeeded for mdev: %s\n",
>>  		     dev_name(mdev_dev(mdev)));
>>  	ret = 0;
>> diff --git a/drivers/gpu/drm/i915/gvt/opregion.c
>> b/drivers/gpu/drm/i915/gvt/opregion.c
>> index 3117991..5c7496d 100644
>> --- a/drivers/gpu/drm/i915/gvt/opregion.c
>> +++ b/drivers/gpu/drm/i915/gvt/opregion.c
>> @@ -114,6 +114,7 @@ void intel_vgpu_clean_opregion(struct intel_vgpu
>> *vgpu)  int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa)
>> {
>>  	int ret;
>> +	unsigned long pfn;
>>
>>  	gvt_dbg_core("vgpu%d: init vgpu opregion\n", vgpu->id);
>>
>> @@ -127,7 +128,12 @@ int intel_vgpu_init_opregion(struct intel_vgpu *vgpu,
>u32 gpa)
>>  		ret = map_vgpu_opregion(vgpu, true);
>>  		if (ret)
>>  			return ret;
>> -	}
>> +	} else {
>> +		pfn = intel_gvt_hypervisor_gfn_to_mfn(vgpu, gpa >>
>PAGE_SHIFT);
>> +		vgpu_opregion(vgpu)->va = memremap(pfn << PAGE_SHIFT,
>> +						INTEL_GVT_OPREGION_SIZE,
>> +						MEMREMAP_WB);
>> +		}
>
>better to use switch for explicit KVM case, and no return check?
Will update and add NULL check.

Thanks for review.
>
>>
>>  	return 0;
>>  }
>> --
>> 2.7.4
>>
>
>--
>Open Source Technology Center, Intel ltd.
>
>$gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827