[PATCH] drm/amd: SVM flag error added at svm_range flags
Felix Kuehling
felix.kuehling at amd.com
Wed May 31 14:46:21 UTC 2023
This flag was never meant for user mode. Alex ended up implementing it
differently in a way that doesn't need to change the flags exposed to
user mode. See his patch "drm/amdkfd: flag added to handle errors from
svm validate and map". This patch is no longer needed.
Regards,
Felix
Am 2023-05-31 um 10:09 schrieb Deucher, Alexander:
>
> [Public]
>
>
> Where is the userspace code which uses this new interface? Can you add
> a link to the relevant branch to the commit message?
>
> Alex
> ------------------------------------------------------------------------
> *From:* amd-gfx <amd-gfx-bounces at lists.freedesktop.org> on behalf of
> Alex Sierra <alex.sierra at amd.com>
> *Sent:* Monday, May 29, 2023 4:09 PM
> *To:* amd-gfx at lists.freedesktop.org <amd-gfx at lists.freedesktop.org>;
> Kuehling, Felix <Felix.Kuehling at amd.com>
> *Cc:* Sierra Guiza, Alejandro (Alex) <Alex.Sierra at amd.com>; Yang,
> Philip <Philip.Yang at amd.com>
> *Subject:* [PATCH] drm/amd: SVM flag error added at svm_range flags
> If a return error is raised during validation and mapping of a
> prange, this flag is set. It is a rare occurrence, but it could happen
> when `amdgpu_hmm_range_get_pages_done` returns true. In such cases,
> the caller should retry. However, it is important to ensure that the
> prange is updated correctly during the retry.
>
> Signed-off-by: Alex Sierra <alex.sierra at amd.com>
> ---
> drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 12 +++++++++++-
> include/uapi/linux/kfd_ioctl.h | 2 ++
> 2 files changed, 13 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> index fcfde9140bce..96abae515bcf 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> @@ -756,10 +756,12 @@ svm_range_apply_attrs(struct kfd_process *p,
> struct svm_range *prange,
> break;
> case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
> *update_mapping = true;
> + attrs[i].value &= ~KFD_IOCTL_SVM_FLAG_ERROR;
> prange->flags |= attrs[i].value;
> break;
> case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
> *update_mapping = true;
> + attrs[i].value &= ~KFD_IOCTL_SVM_FLAG_ERROR;
> prange->flags &= ~attrs[i].value;
> break;
> case KFD_IOCTL_SVM_ATTR_GRANULARITY:
> @@ -778,6 +780,9 @@ svm_range_is_same_attrs(struct kfd_process *p,
> struct svm_range *prange,
> uint32_t i;
> int gpuidx;
>
> + if (prange->flags & KFD_IOCTL_SVM_FLAG_ERROR)
> + return false;
> +
> for (i = 0; i < nattr; i++) {
> switch (attrs[i].type) {
> case KFD_IOCTL_SVM_ATTR_PREFERRED_LOC:
> @@ -1657,8 +1662,11 @@ static int svm_range_validate_and_map(struct
> mm_struct *mm,
> unreserve_out:
> svm_range_unreserve_bos(&ctx);
>
> - if (!r)
> + prange->flags |= KFD_IOCTL_SVM_FLAG_ERROR;
> + if (!r) {
> + prange->flags &= ~KFD_IOCTL_SVM_FLAG_ERROR;
> prange->validate_timestamp = ktime_get_boottime();
> + }
>
> return r;
> }
> @@ -3674,9 +3682,11 @@ svm_range_get_attr(struct kfd_process *p,
> struct mm_struct *mm,
> break;
> case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
> attrs[i].value = flags_and;
> + attrs[i].value &= ~KFD_IOCTL_SVM_FLAG_ERROR;
> break;
> case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
> attrs[i].value = ~flags_or;
> + attrs[i].value &= ~KFD_IOCTL_SVM_FLAG_ERROR;
> break;
> case KFD_IOCTL_SVM_ATTR_GRANULARITY:
> attrs[i].value = (uint32_t)granularity;
> diff --git a/include/uapi/linux/kfd_ioctl.h
> b/include/uapi/linux/kfd_ioctl.h
> index 7e19a2d1e907..2b6f68bd06da 100644
> --- a/include/uapi/linux/kfd_ioctl.h
> +++ b/include/uapi/linux/kfd_ioctl.h
> @@ -1125,6 +1125,8 @@ struct kfd_ioctl_cross_memory_copy_deprecated_args {
> #define KFD_IOCTL_SVM_FLAG_GPU_READ_MOSTLY 0x00000020
> /* Keep GPU memory mapping always valid as if XNACK is disable */
> #define KFD_IOCTL_SVM_FLAG_GPU_ALWAYS_MAPPED 0x00000040
> +/* Set during svm validation and map error */
> +#define KFD_IOCTL_SVM_FLAG_ERROR 0x80000000
>
> /**
> * kfd_ioctl_svm_op - SVM ioctl operations
> --
> 2.32.0
>
More information about the amd-gfx
mailing list