AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Felix Kuehling <felix.kuehling@amd.com>
To: "Christian König" <ckoenig.leichtzumerken@gmail.com>,
	"Philip Yang" <Philip.Yang@amd.com>,
	amd-gfx@lists.freedesktop.org
Subject: Re: [PATCH v2 2/2] drm/amdkfd: heavy-weight flush TLB after unmap
Date: Fri, 14 May 2021 12:12:53 -0400	[thread overview]
Message-ID: <fedfd1b3-3f18-dda0-c189-fd7a0675beee@amd.com> (raw)
In-Reply-To: <d5fba451-f75b-b77c-7fbf-a289914e97e5@gmail.com>

Am 2021-05-14 um 11:14 a.m. schrieb Christian König:
> Am 14.05.21 um 16:42 schrieb Philip Yang:
>> Need do a heavy-weight TLB flush to make sure we have no more dirty data
>> in the cache for the unmapped pages.
>>
>> Define enum TLB_FLUSH_TYPE, add flush_type parameter to
>> amdgpu_amdkfd_flush_gpu_tlb_pasid.
>>
>> Signed-off-by: Philip Yang <Philip.Yang@amd.com>
>
> Reviewed-by: Christian König <christian.koenig@amd.com>

Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>


>
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 4 ++--
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 9 ++++++++-
>>   drivers/gpu/drm/amd/amdkfd/kfd_process.c   | 2 +-
>>   drivers/gpu/drm/amd/amdkfd/kfd_svm.c       | 4 ++--
>>   4 files changed, 13 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
>> index 5f6696a3c778..bfab2f9fdd17 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
>> @@ -670,10 +670,10 @@ int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct
>> kgd_dev *kgd, uint16_t vmid)
>>       return 0;
>>   }
>>   -int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd,
>> uint16_t pasid)
>> +int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t
>> pasid,
>> +                      enum TLB_FLUSH_TYPE flush_type)
>>   {
>>       struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
>> -    const uint32_t flush_type = 0;
>>       bool all_hub = false;
>>         if (adev->family == AMDGPU_FAMILY_AI)
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
>> index 313ee49b9f17..f4de50de359b 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
>> @@ -36,6 +36,12 @@
>>     extern uint64_t amdgpu_amdkfd_total_mem_size;
>>   +enum TLB_FLUSH_TYPE {
>> +    TLB_FLUSH_LEGACY = 0,
>> +    TLB_FLUSH_LIGHTWEIGHT,
>> +    TLB_FLUSH_HEAVYWEIGHT
>> +};
>> +
>>   struct amdgpu_device;
>>     struct kfd_bo_va_list {
>> @@ -135,7 +141,8 @@ int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd,
>> enum kgd_engine_type engine,
>>   void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle);
>>   bool amdgpu_amdkfd_have_atomics_support(struct kgd_dev *kgd);
>>   int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t
>> vmid);
>> -int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t
>> pasid);
>> +int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t
>> pasid,
>> +                      enum TLB_FLUSH_TYPE flush_type);
>>     bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32
>> vmid);
>>   diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> index 9d4f527bda7c..5b6c5669c03d 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
>> @@ -1853,7 +1853,7 @@ void kfd_flush_tlb(struct kfd_process_device *pdd)
>>                               pdd->qpd.vmid);
>>       } else {
>>           amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->kgd,
>> -                        pdd->process->pasid);
>> +                    pdd->process->pasid, TLB_FLUSH_LEGACY);
>>       }
>>   }
>>   diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
>> b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
>> index a52e4594785f..f2901e7a993e 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
>> @@ -1127,7 +1127,7 @@ svm_range_unmap_from_gpus(struct svm_range
>> *prange, unsigned long start,
>>                   break;
>>           }
>>           amdgpu_amdkfd_flush_gpu_tlb_pasid((struct kgd_dev *)adev,
>> -                          p->pasid);
>> +                    p->pasid, TLB_FLUSH_HEAVYWEIGHT);
>>       }
>>         return r;
>> @@ -1183,7 +1183,7 @@ svm_range_map_to_gpu(struct amdgpu_device
>> *adev, struct amdgpu_vm *vm,
>>             p = container_of(prange->svms, struct kfd_process, svms);
>>           amdgpu_amdkfd_flush_gpu_tlb_pasid((struct kgd_dev *)adev,
>> -                          p->pasid);
>> +                        p->pasid, TLB_FLUSH_LEGACY);
>>       }
>>   out:
>>       prange->mapping.bo_va = NULL;
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

      reply	other threads:[~2021-05-14 16:13 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-05-13 16:58 [PATCH 1/2] Revert "drm/amdkfd: flush TLB after updating GPU page table" Philip Yang
2021-05-13 16:58 ` [PATCH 2/2] drm/amdkfd: heavy-weight flush TLB after unmap Philip Yang
2021-05-13 17:18   ` Felix Kuehling
2021-05-14  7:24   ` Christian König
2021-05-14 14:42   ` [PATCH v2 " Philip Yang
2021-05-14 15:14     ` Christian König
2021-05-14 16:12       ` Felix Kuehling [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=fedfd1b3-3f18-dda0-c189-fd7a0675beee@amd.com \
    --to=felix.kuehling@amd.com \
    --cc=Philip.Yang@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=ckoenig.leichtzumerken@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox