> > + r = virt->ops->rel_full_gpu(adev, init); > + adev->virt.caps |= AMDGPU_SRIOV_CAPS_RUNTIME; > > [ml] what about rel_full_gpu() failed ? and if rel_full_gpu is > designed not allowed to fail then you should type it as non-return > function. ( I remember in my first initial implementation on rel_gpu > function it is is non-return > function) I think it is more flexible, we can handle error according to return result. [ML] then please handle the error accordingly, but you always do "|= AMDGPU_SRIOV_CAPS_RUNTIME" no matter above rel_full_gpu() return okay or not , isn't it ? That RUNTIME flag should only be set after successfully release the full access mode on GPU... -----Original Message----- From: Yu, Xiangliang Sent: Wednesday, January 11, 2017 10:27 PM To: Liu, Monk <Monk.Liu at amd.com>; amd-gfx at lists.freedesktop.org Subject: RE: [V3 05/11] drm/amdgpu/virt: add high level interfaces for virt > This patch is not derived from my work (amd-sriov-4.3/4.6), please > don't add my "signed-off-by: Monk Liu " line on it, Seems you add an > upper layer interface to invoke on gpu_request/release routines, And I > will also take role in the reviewing of it: > > + adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME; > + return virt->ops->req_full_gpu(adev, init); > > [ml]One nitpick , you need place " adev->virt.caps &= > ~AMDGPU_SRIOV_CAPS_RUNTIME;" in the second line, I think it make more clear as it is not runtime when entering full access mode. > > + r = virt->ops->rel_full_gpu(adev, init); > + adev->virt.caps |= AMDGPU_SRIOV_CAPS_RUNTIME; > > [ml] what about rel_full_gpu() failed ? and if rel_full_gpu is > designed not allowed to fail then you should type it as non-return > function. ( I remember in my first initial implementation on rel_gpu > function it is is non-return > function) I think it is more flexible, we can handle error according to return result. > > > + adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME; > + return virt->ops->reset_gpu(adev); > + } > > [ML] " adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME" should be put > to the second line as well, like: > > r = virt-ops->reset_gpu(adev); > If (!r) > adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME; else > BUG(); /* we shouldn't fail in requesting GPU_RESET, check > hypervisor please */ return r; I think we can't hang out VM so rudely, need to print out error message and do error handling to enhance the process later. > > BR Monk > > -----Original Message----- > From: Xiangliang Yu [mailto:Xiangliang.Yu at amd.com] > Sent: Wednesday, January 11, 2017 9:18 PM > To: amd-gfx at lists.freedesktop.org > Cc: Yu, Xiangliang <Xiangliang.Yu at amd.com>; Liu, Monk > <Monk.Liu at amd.com> > Subject: [V3 05/11] drm/amdgpu/virt: add high level interfaces for > virt > > Add high level interfaces that is not relate to specific asic. So asic > files just need to implement the interfaces to support virtualization. > > Signed-off-by: Xiangliang Yu <Xiangliang.Yu at amd.com> > Signed-off-by: Monk Liu <Monk.Liu at amd.com> > --- > drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c | 57 > ++++++++++++++++++++++++++++++++ > drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 14 ++++++++ > 2 files changed, 71 insertions(+) > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c > b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c > index 7861b6b..4e7df8e 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c > @@ -80,3 +80,60 @@ void amdgpu_virt_kiq_wreg(struct amdgpu_device > *adev, uint32_t reg, uint32_t v) > DRM_ERROR("wait for kiq fence error: %ld.\n", r); > fence_put(f); > } > + > +/** > + * amdgpu_virt_request_full_gpu() - request full gpu access > + * @amdgpu: amdgpu device. > + * @init: is driver init time. > + * When start to init/fini driver, first need to request full gpu access. > + * Return: Zero if request success, otherwise will return error. > + */ > +int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool > +init) { > + struct amdgpu_virt *virt = &adev->virt; > + > + if (virt->ops && virt->ops->req_full_gpu) { > + adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME; > + return virt->ops->req_full_gpu(adev, init); > + } > + > + return 0; > +} > + > +/** > + * amdgpu_virt_release_full_gpu() - release full gpu access > + * @amdgpu: amdgpu device. > + * @init: is driver init time. > + * When finishing driver init/fini, need to release full gpu access. > + * Return: Zero if release success, otherwise will returen error. > + */ > +int amdgpu_virt_release_full_gpu(struct amdgpu_device *adev, bool > +init) { > + struct amdgpu_virt *virt = &adev->virt; > + int r; > + > + if (virt->ops && virt->ops->rel_full_gpu) { > + r = virt->ops->rel_full_gpu(adev, init); > + adev->virt.caps |= AMDGPU_SRIOV_CAPS_RUNTIME; > + return r; > + } > + return 0; > +} > + > +/** > + * amdgpu_virt_reset_gpu() - reset gpu > + * @amdgpu: amdgpu device. > + * Send reset command to GPU hypervisor to reset GPU that VM is using > + * Return: Zero if reset success, otherwise will return error. > + */ > +int amdgpu_virt_reset_gpu(struct amdgpu_device *adev) { > + struct amdgpu_virt *virt = &adev->virt; > + > + if (virt->ops && virt->ops->reset_gpu) { > + adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME; > + return virt->ops->reset_gpu(adev); > + } > + > + return 0; > +} > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h > b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h > index 2869980..45771b0 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h > @@ -29,11 +29,22 @@ > #define AMDGPU_SRIOV_CAPS_IS_VF (1 << 2) /* this GPU is a virtual > function */ > #define AMDGPU_PASSTHROUGH_MODE (1 << 3) /* thw whole GPU is > pass through for VM */ > #define AMDGPU_SRIOV_CAPS_RUNTIME (1 << 4) /* is out of full access > mode */ > + > +/** > + * struct amdgpu_virt_ops - amdgpu device virt operations */ struct > +amdgpu_virt_ops { > + int (*req_full_gpu)(struct amdgpu_device *adev, bool init); > + int (*rel_full_gpu)(struct amdgpu_device *adev, bool init); > + int (*reset_gpu)(struct amdgpu_device *adev); }; > + > /* GPU virtualization */ > struct amdgpu_virt { > uint32_t caps; > uint32_t reg_val_offs; > struct mutex lock; > + const struct amdgpu_virt_ops *ops; > }; > > #define amdgpu_sriov_enabled(adev) \ > @@ -63,5 +74,8 @@ static inline bool is_virtual_machine(void) void > amdgpu_virt_init_setting(struct amdgpu_device *adev); uint32_t > amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg); void > amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, > uint32_t v); > +int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool > +init); int amdgpu_virt_release_full_gpu(struct amdgpu_device *adev, > +bool init); int amdgpu_virt_reset_gpu(struct amdgpu_device *adev); > > #endif > -- > 2.7.4
