[Intel-gfx] [PATCH v4 6/6] drm/i915: expose rcs topology through query uAPI

Lionel Landwerlin lionel.g.landwerlin at intel.com
Mon Jan 15 18:23:32 UTC 2018


On 15/01/18 17:54, Tvrtko Ursulin wrote:
>
> On 15/01/2018 14:41, Lionel Landwerlin wrote:
>> With the introduction of asymmetric slices in CNL, we cannot rely on
>> the previous SUBSLICE_MASK getparam to tell userspace what subslices
>> are available. Here we introduce a more detailed way of querying the
>> Gen's GPU topology that doesn't aggregate numbers.
>>
>> This is essential for monitoring parts of the GPU with the OA unit,
>> because counters need to be normalized to the number of
>> EUs/subslices/slices. The current aggregated numbers like EU_TOTAL do
>> not gives us sufficient information.
>>
>> As a bonus we can draw representations of the GPU :
>>
>>          https://imgur.com/a/vuqpa
>>
>> v2: Rename uapi struct s/_mask/_info/ (Tvrtko)
>>      Report max_slice/subslice/eus_per_subslice rather than strides 
>> (Tvrtko)
>>      Add uapi macros to read data from *_info structs (Tvrtko)
>>
>> v3: Use !!(v & DRM_I915_BIT()) for uapi macros instead of custom 
>> shifts (Tvrtko)
>>
>> Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin at intel.com>
>> ---
>>   drivers/gpu/drm/i915/i915_query.c | 134 
>> ++++++++++++++++++++++++++++++++++++++
>>   include/uapi/drm/i915_drm.h       |  53 +++++++++++++++
>>   2 files changed, 187 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/i915/i915_query.c 
>> b/drivers/gpu/drm/i915/i915_query.c
>> index 5694cfea4553..465ec18a472f 100644
>> --- a/drivers/gpu/drm/i915/i915_query.c
>> +++ b/drivers/gpu/drm/i915/i915_query.c
>> @@ -25,8 +25,129 @@
>>   #include "i915_drv.h"
>>   #include <uapi/drm/i915_drm.h>
>>   +static int query_slices_info(struct drm_i915_private *dev_priv,
>> +                 struct drm_i915_query_item *query_item)
>> +{
>> +    const struct sseu_dev_info *sseu = &INTEL_INFO(dev_priv)->sseu;
>> +    struct drm_i915_query_slices_info slices_info;
>> +    u32 data_length, length;
>> +
>> +    if (sseu->max_slices == 0)
>> +        return -ENODEV;
>> +
>> +    data_length = sizeof(sseu->slice_mask);
>> +    length = sizeof(slices_info) + data_length;
>> +
>> +    /*
>> +     * If we ever change the internal slice mask data type, we'll 
>> need to
>> +     * update this function.
>> +     */
>> +    BUILD_BUG_ON(sizeof(u8) != sizeof(sseu->slice_mask));
>> +
>> +    if (query_item->length == 0) {
>> +        query_item->length = length;
>> +        return 0;
>> +    }
>> +
>> +    if (query_item->length != length)
>> +        return -EINVAL;
>> +
>> +    memset(&slices_info, 0, sizeof(slices_info));
>> +    slices_info.max_slices = sseu->max_slices;
>> +
>> +    if (copy_to_user(u64_to_user_ptr(query_item->data_ptr), 
>> &slices_info,
>> +             sizeof(slices_info)))
>> +        return -EFAULT;
>> +
>> +    if (copy_to_user(u64_to_user_ptr(query_item->data_ptr +
>> +                     offsetof(struct drm_i915_query_slices_info, 
>> data)),
>> +             &sseu->slice_mask, data_length))
>> +        return -EFAULT;
>> +
>> +    return 0;
>> +}
>> +
>> +static int query_subslices_info(struct drm_i915_private *dev_priv,
>> +                struct drm_i915_query_item *query_item)
>> +{
>> +    const struct sseu_dev_info *sseu = &INTEL_INFO(dev_priv)->sseu;
>> +    struct drm_i915_query_subslices_info subslices_info;
>> +    u32 data_length, length;
>> +
>> +    if (sseu->max_slices == 0)
>> +        return -ENODEV;
>> +
>> +    memset(&subslices_info, 0, sizeof(subslices_info));
>> +    subslices_info.max_slices = sseu->max_slices;
>> +    subslices_info.max_subslices = sseu->max_subslices;
>> +
>> +    data_length = subslices_info.max_slices *
>> +        DIV_ROUND_UP(subslices_info.max_subslices,
>> +                 sizeof(sseu->subslice_mask[0]) * BITS_PER_BYTE);
>> +    length = sizeof(subslices_info) + data_length;
>> +
>> +    if (query_item->length == 0) {
>> +        query_item->length = length;
>> +        return 0;
>> +    }
>> +
>> +    if (query_item->length != length)
>> +        return -EINVAL;
>> +
>> +    if (copy_to_user(u64_to_user_ptr(query_item->data_ptr), 
>> &subslices_info,
>> +             sizeof(subslices_info)))
>> +        return -EFAULT;
>> +
>> +    if (copy_to_user(u64_to_user_ptr(query_item->data_ptr +
>> +                     offsetof(struct drm_i915_query_subslices_info, 
>> data)),
>> +             sseu->subslice_mask, data_length))
>> +        return -EFAULT;
>> +
>> +    return 0;
>> +}
>> +
>> +static int query_eus_info(struct drm_i915_private *dev_priv,
>> +              struct drm_i915_query_item *query_item)
>> +{
>> +    const struct sseu_dev_info *sseu = &INTEL_INFO(dev_priv)->sseu;
>> +    struct drm_i915_query_eus_info eus_info;
>> +    u32 data_length, length;
>> +
>> +    if (sseu->max_slices == 0)
>> +        return -ENODEV;
>> +
>> +    memset(&eus_info, 0, sizeof(eus_info));
>> +    eus_info.max_slices = sseu->max_slices;
>> +    eus_info.max_subslices = sseu->max_subslices;
>> +    eus_info.max_eus_per_subslice = sseu->max_eus_per_subslice;
>> +
>> +    data_length = eus_info.max_slices * eus_info.max_subslices *
>> +        DIV_ROUND_UP(eus_info.max_eus_per_subslice, BITS_PER_BYTE);
>> +    length = sizeof(eus_info) + data_length;
>> +
>> +    if (query_item->length == 0) {
>> +        query_item->length = length;
>> +        return 0;
>> +    }
>> +
>> +    if (query_item->length != length)
>> +        return -EINVAL;
>> +
>> +    if (copy_to_user(u64_to_user_ptr(query_item->data_ptr), &eus_info,
>> +             sizeof(eus_info)))
>> +        return -EFAULT;
>> +
>> +    if (copy_to_user(u64_to_user_ptr(query_item->data_ptr +
>> +                     offsetof(struct drm_i915_query_eus_info, data)),
>> +             sseu->eu_mask, data_length))
>> +        return -EFAULT;
>
> I think in all three queries, once you have the length, you could 
> extract the four if statement above into a common helper? Arguments 
> are query_item, length, data_length, ptr_query_item, ptr_data, and 
> offset_data, unless I am missing something. Up to you if you would 
> find that a worthy compaction.

I just tried that but the result doesn't really look good.
You need to maintain the control flow via an indirect return value...

>
>> +
>> +    return 0;
>> +}
>> +
>>   int i915_query_ioctl(struct drm_device *dev, void *data, struct 
>> drm_file *file)
>>   {
>> +    struct drm_i915_private *dev_priv = to_i915(dev);
>>       struct drm_i915_query *args = data;
>>       struct drm_i915_query_item __user *user_item_ptr =
>>           u64_to_user_ptr(args->items_ptr);
>> @@ -34,15 +155,28 @@ int i915_query_ioctl(struct drm_device *dev, 
>> void *data, struct drm_file *file)
>>         for (i = 0; i < args->num_items; i++, user_item_ptr++) {
>>           struct drm_i915_query_item item;
>> +        int ret;
>>             if (copy_from_user(&item, user_item_ptr, sizeof(item)))
>>               return -EFAULT;
>>             switch (item.query_id) {
>> +        case DRM_I915_QUERY_ID_SLICES_INFO:
>> +            ret = query_slices_info(dev_priv, &item);
>> +            break;
>> +        case DRM_I915_QUERY_ID_SUBSLICES_INFO:
>> +            ret = query_subslices_info(dev_priv, &item);
>> +            break;
>> +        case DRM_I915_QUERY_ID_EUS_INFO:
>> +            ret = query_eus_info(dev_priv, &item);
>> +            break;
>>           default:
>>               return -EINVAL;
>>           }
>>   +        if (ret)
>> +            return ret;
>> +
>>           if (copy_to_user(user_item_ptr, &item, sizeof(item)))
>>               return -EFAULT;
>>       }
>> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
>> index 39e93f10f2cd..d25f21e3c107 100644
>> --- a/include/uapi/drm/i915_drm.h
>> +++ b/include/uapi/drm/i915_drm.h
>> @@ -1618,6 +1618,9 @@ struct drm_i915_perf_oa_config {
>>     struct drm_i915_query_item {
>>       __u64 query_id;
>> +#define DRM_I915_QUERY_ID_SLICES_INFO    0x01
>> +#define DRM_I915_QUERY_ID_SUBSLICES_INFO 0x02
>> +#define DRM_I915_QUERY_ID_EUS_INFO       0x03
>
> Would DRM_I915_QUERY_ID_"singular"_INFO read better? In struct names 
> as well if so?

Hmm... Sorry, I don't understand your comment here.
What would be "singular" ?

>
>>         /*
>>        * When set to zero by userspace, this is filled with the size 
>> of the
>> @@ -1644,6 +1647,56 @@ struct drm_i915_query {
>>       __u64 items_ptr;
>>   };
>
> Okay.. up to here it's all pretty much fine.
>
>>   +#define DRM_I915_BIT(bit) (1 << (bit))
>> +
>> +/* Data written by the kernel with query 
>> DRM_I915_QUERY_ID_SLICES_INFO :
>> + *
>> + * data: each bit indicates whether a slice is available (1) or 
>> fused off (0).
>> + *       Use DRM_I915_QUERY_SLICE_AVAILABLE() to query a given slice's
>> + *       availability.
>> + */
>> +struct drm_i915_query_slices_info {
>> +    __u32 max_slices;
>> +
>> +#define DRM_I915_QUERY_SLICE_AVAILABLE(info, slice) \
>> +    !!((info)->data[(slice) / 8] & DRM_I915_BIT((slice) % 8))
>> +    __u8 data[];
>> +};
>> +
>> +/* Data written by the kernel with query 
>> DRM_I915_QUERY_ID_SUBSLICES_INFO :
>> + *
>> + * data: each bit indicates whether a subslice is available (1) or 
>> fused off
>> + *       (0). Use DRM_I915_QUERY_SUBSLICE_AVAILABLE() to query a given
>> + *       subslice's availability.
>> + */
>> +struct drm_i915_query_subslices_info {
>> +    __u32 max_slices;
>> +    __u32 max_subslices;
>> +
>> +#define DRM_I915_QUERY_SUBSLICE_AVAILABLE(info, slice, subslice) \
>> +    !!((info)->data[(slice) * ALIGN((info)->max_subslices, 8) / 8 + \
>> +            (subslice) / 8] & DRM_I915_BIT((subslice) % 8))
>> +    __u8 data[];
>> +};
>> +
>> +/* Data written by the kernel with query DRM_I915_QUERY_ID_EUS_INFO :
>> + *
>> + * data: Each bit indicates whether a subslice is available (1) or 
>> fused off
>> + *       (0). Use DRM_I915_QUERY_EU_AVAILABLE() to query a given EU's
>> + *       availability.
>> + */
>> +struct drm_i915_query_eus_info {
>> +    __u32 max_slices;
>> +    __u32 max_subslices;
>> +    __u32 max_eus_per_subslice;
>> +
>> +#define DRM_I915_QUERY_EU_AVAILABLE(info, slice, subslice, eu) \
>> +    !!((info)->data[(slice) * ALIGN((info)->max_eus_per_subslice, 8) 
>> / 8 * (info)->max_subslices + \
>> +            (subslice) * ALIGN((info)->max_eus_per_subslice, 8) / 8 + \
>> +            (eu) / 8] & DRM_I915_BIT((eu) % 8))
>> +    __u8 data[];
>> +};
>> +
>>   #if defined(__cplusplus)
>>   }
>>   #endif
>>
>
> This last bit however I'm a bit unsure of.
>
> I know I suggested the accessor macros but now they look a bit 
> unsightly. I don't have ideas on how to improve them though.
>
> And the unspecified length array is also potentially questionable.
>
> I'll take the liberty of Cc-ing Chris and Joonas to start with to see 
> if they have comments on the above.
>
> Regards,
>
> Tvrtko
>
Thanks again for your review.




More information about the Intel-gfx mailing list