From: "Christian König" <christian.koenig@amd.com>
To: Prike Liang <Prike.Liang@amd.com>, amd-gfx@lists.freedesktop.org
Cc: Alexander.Deucher@amd.com, felix.kuehling@amd.com,
Harish.Kasiviswanathan@amd.com
Subject: Re: [PATCH] drm/amdkfd: correct the SVM DMA device unmap direction
Date: Tue, 5 Nov 2024 12:04:14 +0100 [thread overview]
Message-ID: <04d4ab32-45a1-4b88-86ee-fb0f35a0ca40@amd.com> (raw)
In-Reply-To: <20241105023332.172404-1-Prike.Liang@amd.com>
Am 05.11.24 um 03:33 schrieb Prike Liang:
> The SVM DMA device unmap direction should be same as
> the DMA map process.
At least of hand that looks like it's only papering over a major problem.
Why are DMA ranges for SVM mapped with a direction in the first place?
That is usually not something we should do.
Regards,
Christian.
>
> Signed-off-by: Prike Liang <Prike.Liang@amd.com>
> ---
> drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 4 ++--
> drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 6 +++---
> drivers/gpu/drm/amd/amdkfd/kfd_svm.h | 3 ++-
> 3 files changed, 7 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> index eacfeb32f35d..9d83bb9dd004 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> @@ -445,7 +445,7 @@ svm_migrate_vma_to_vram(struct kfd_node *node, struct svm_range *prange,
> pr_debug("successful/cpages/npages 0x%lx/0x%lx/0x%lx\n",
> mpages, cpages, migrate.npages);
>
> - svm_range_dma_unmap_dev(adev->dev, scratch, 0, npages);
> + svm_range_dma_unmap_dev(adev->dev, scratch, 0, npages, DMA_TO_DEVICE);
>
> out_free:
> kvfree(buf);
> @@ -750,7 +750,7 @@ svm_migrate_vma_to_ram(struct kfd_node *node, struct svm_range *prange,
> svm_migrate_copy_done(adev, mfence);
> migrate_vma_finalize(&migrate);
>
> - svm_range_dma_unmap_dev(adev->dev, scratch, 0, npages);
> + svm_range_dma_unmap_dev(adev->dev, scratch, 0, npages, DMA_FROM_DEVICE);
>
> out_free:
> kvfree(buf);
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> index 3e2911895c74..c21485fe6cbb 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> @@ -233,9 +233,9 @@ svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
> }
>
> void svm_range_dma_unmap_dev(struct device *dev, dma_addr_t *dma_addr,
> - unsigned long offset, unsigned long npages)
> + unsigned long offset, unsigned long npages,
> + enum dma_data_direction dir)
> {
> - enum dma_data_direction dir = DMA_BIDIRECTIONAL;
> int i;
>
> if (!dma_addr)
> @@ -272,7 +272,7 @@ void svm_range_dma_unmap(struct svm_range *prange)
> }
> dev = &pdd->dev->adev->pdev->dev;
>
> - svm_range_dma_unmap_dev(dev, dma_addr, 0, prange->npages);
> + svm_range_dma_unmap_dev(dev, dma_addr, 0, prange->npages, DMA_BIDIRECTIONAL);
> }
> }
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
> index bddd24f04669..5370d68bc5b2 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
> @@ -182,7 +182,8 @@ void svm_range_add_list_work(struct svm_range_list *svms,
> enum svm_work_list_ops op);
> void schedule_deferred_list_work(struct svm_range_list *svms);
> void svm_range_dma_unmap_dev(struct device *dev, dma_addr_t *dma_addr,
> - unsigned long offset, unsigned long npages);
> + unsigned long offset, unsigned long npages,
> + enum dma_data_direction dir);
> void svm_range_dma_unmap(struct svm_range *prange);
> int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
> uint64_t *svm_priv_data_size);
next prev parent reply other threads:[~2024-11-05 11:04 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-11-05 2:33 [PATCH] drm/amdkfd: correct the SVM DMA device unmap direction Prike Liang
2024-11-05 11:04 ` Christian König [this message]
2024-11-05 16:34 ` Felix Kuehling
2024-11-06 12:24 ` Christian König
2024-11-08 9:15 ` Liang, Prike
2024-11-08 9:39 ` Christian König
2024-11-11 3:06 ` Liang, Prike
2024-11-12 13:42 ` Christian König
2024-12-02 7:32 ` Liang, Prike
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=04d4ab32-45a1-4b88-86ee-fb0f35a0ca40@amd.com \
--to=christian.koenig@amd.com \
--cc=Alexander.Deucher@amd.com \
--cc=Harish.Kasiviswanathan@amd.com \
--cc=Prike.Liang@amd.com \
--cc=amd-gfx@lists.freedesktop.org \
--cc=felix.kuehling@amd.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