[PATCH 2/2] drm/amdkfd: Relocate TBA/TMA to opposite side of VM hole (v2)

Christian König christian.koenig at amd.com
Wed Jan 31 13:38:45 UTC 2024



Am 30.01.24 um 21:08 schrieb Felix Kuehling:
> The TBA and TMA, along with an unused IB allocation, reside at low
> addresses in the VM address space. A stray VM fault which hits these
> pages must be serviced by making their page table entries invalid.
> The scheduler depends upon these pages being resident and fails,
> preventing a debugger from inspecting the failure state.
>
> By relocating these pages above 47 bits in the VM address space they
> can only be reached when bits [63:48] are set to 1. This makes it much
> less likely for a misbehaving program to generate accesses to them.
> The current placement at VA (PAGE_SIZE*2) is readily hit by a NULL
> access with a small offset.
>
> v2:
> - Move it to the reserved space to avoid concflicts with Mesa
> - Add macros to make reserved space management easier
>
> Cc: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam at amd.com>
> Cc: Christian Koenig <christian.koenig at amd.com>
> Signed-off-by: Jay Cornwall <jay.cornwall at amd.com>
> Signed-off-by: Felix Kuehling <felix.kuehling at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c      |  4 +--
>   drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c    |  7 ++---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h       | 10 ++++++-
>   drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c | 30 +++++++++++---------
>   4 files changed, 29 insertions(+), 22 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
> index 823d31f4a2a3..53d0a458d78e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
> @@ -28,9 +28,9 @@
>   
>   uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev)
>   {
> -	uint64_t addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT;
> +	uint64_t addr = AMDGPU_VA_RESERVED_CSA_START(
> +		adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT);

Maybe move the "adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT" part 
into the macro. That should be identical for all use cases.

Apart from that looks really good to me.

Regards,
Christian.

>   
> -	addr -= AMDGPU_VA_RESERVED_CSA_SIZE;
>   	addr = amdgpu_gmc_sign_extend(addr);
>   
>   	return addr;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
> index 3d0d56087d41..9e769ef50f2e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
> @@ -45,11 +45,8 @@
>    */
>   static inline u64 amdgpu_seq64_get_va_base(struct amdgpu_device *adev)
>   {
> -	u64 addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT;
> -
> -	addr -= AMDGPU_VA_RESERVED_TOP;
> -
> -	return addr;
> +	return AMDGPU_VA_RESERVED_SEQ64_START(
> +		adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT);
>   }
>   
>   /**
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> index 2c4053b29bb3..c2407f6a7e83 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
> @@ -137,9 +137,17 @@ struct amdgpu_mem_stats;
>   
>   /* Reserve space at top/bottom of address space for kernel use */
>   #define AMDGPU_VA_RESERVED_CSA_SIZE		(2ULL << 20)
> +#define AMDGPU_VA_RESERVED_CSA_START(top)	((top) \
> +						 - AMDGPU_VA_RESERVED_CSA_SIZE)
>   #define AMDGPU_VA_RESERVED_SEQ64_SIZE		(2ULL << 20)
> +#define AMDGPU_VA_RESERVED_SEQ64_START(top)	(AMDGPU_VA_RESERVED_CSA_START(top) \
> +						 - AMDGPU_VA_RESERVED_SEQ64_SIZE)
> +#define AMDGPU_VA_RESERVED_TRAP_SIZE		(2ULL << 12)
> +#define AMDGPU_VA_RESERVED_TRAP_START(top)	(AMDGPU_VA_RESERVED_SEQ64_START(top) \
> +						 - AMDGPU_VA_RESERVED_TRAP_SIZE)
>   #define AMDGPU_VA_RESERVED_BOTTOM		(1ULL << 16)
> -#define AMDGPU_VA_RESERVED_TOP			(AMDGPU_VA_RESERVED_SEQ64_SIZE + \
> +#define AMDGPU_VA_RESERVED_TOP			(AMDGPU_VA_RESERVED_TRAP_SIZE + \
> +						 AMDGPU_VA_RESERVED_SEQ64_SIZE + \
>   						 AMDGPU_VA_RESERVED_CSA_SIZE)
>   
>   /* See vm_update_mode */
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
> index 6604a3f99c5e..f899cce25b2a 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
> @@ -36,6 +36,7 @@
>   #include <linux/mm.h>
>   #include <linux/mman.h>
>   #include <linux/processor.h>
> +#include "amdgpu_vm.h"
>   
>   /*
>    * The primary memory I/O features being added for revisions of gfxip
> @@ -326,10 +327,16 @@ static void kfd_init_apertures_vi(struct kfd_process_device *pdd, uint8_t id)
>   	 * with small reserved space for kernel.
>   	 * Set them to CANONICAL addresses.
>   	 */
> -	pdd->gpuvm_base = SVM_USER_BASE;
> +	pdd->gpuvm_base = max(SVM_USER_BASE, AMDGPU_VA_RESERVED_BOTTOM);
>   	pdd->gpuvm_limit =
>   		pdd->dev->kfd->shared_resources.gpuvm_size - 1;
>   
> +	/* dGPUs: the reserved space for kernel
> +	 * before SVM
> +	 */
> +	pdd->qpd.cwsr_base = SVM_CWSR_BASE;
> +	pdd->qpd.ib_base = SVM_IB_BASE;
> +
>   	pdd->scratch_base = MAKE_SCRATCH_APP_BASE_VI();
>   	pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base);
>   }
> @@ -339,18 +346,19 @@ static void kfd_init_apertures_v9(struct kfd_process_device *pdd, uint8_t id)
>   	pdd->lds_base = MAKE_LDS_APP_BASE_V9();
>   	pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base);
>   
> -        /* Raven needs SVM to support graphic handle, etc. Leave the small
> -         * reserved space before SVM on Raven as well, even though we don't
> -         * have to.
> -         * Set gpuvm_base and gpuvm_limit to CANONICAL addresses so that they
> -         * are used in Thunk to reserve SVM.
> -         */
> -        pdd->gpuvm_base = SVM_USER_BASE;
> +	pdd->gpuvm_base = AMDGPU_VA_RESERVED_BOTTOM;
>   	pdd->gpuvm_limit =
>   		pdd->dev->kfd->shared_resources.gpuvm_size - 1;
>   
>   	pdd->scratch_base = MAKE_SCRATCH_APP_BASE_V9();
>   	pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base);
> +
> +	/*
> +	 * Place TBA/TMA on opposite side of VM hole to prevent
> +	 * stray faults from triggering SVM on these pages.
> +	 */
> +	pdd->qpd.cwsr_base = AMDGPU_VA_RESERVED_TRAP_START(
> +		pdd->dev->adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT);
>   }
>   
>   int kfd_init_apertures(struct kfd_process *process)
> @@ -407,12 +415,6 @@ int kfd_init_apertures(struct kfd_process *process)
>   					return -EINVAL;
>   				}
>   			}
> -
> -                        /* dGPUs: the reserved space for kernel
> -                         * before SVM
> -                         */
> -                        pdd->qpd.cwsr_base = SVM_CWSR_BASE;
> -                        pdd->qpd.ib_base = SVM_IB_BASE;
>   		}
>   
>   		dev_dbg(kfd_device, "node id %u\n", id);



More information about the amd-gfx mailing list