hw_base_encode: refactor picture allocation/freeing

This commit cleans up and refactors the mess of private state upon
private state that used to be.

Now, FFHWBaseEncodePicture is fully initialized upon call-time,
and, most importantly, this lets APIs which require initialization
data for frames (VkImageViews) to initialize this for both the
input image, and the reconstruction (DPB) image.

Signed-off-by: Tong Wu <wutong1208@outlook.com>

14 files changed, 201 insertions, 211 deletions

diff --git a/libavcodec/d3d12va_encode.c b/libavcodec/d3d12va_encode.c
index 9ee9da41e3..d437750598 100644
--- a/libavcodec/d3d12va_encode.c
+++ b/libavcodec/d3d12va_encode.c
@@ -107,10 +107,10 @@ static int d3d12va_discard_command_allocator(AVCodecContext *avctx, ID3D12Comman
 }
 
 static int d3d12va_encode_wait(AVCodecContext *avctx,
-                               D3D12VAEncodePicture *pic)
+                               FFHWBaseEncodePicture *base_pic)
 {
-    D3D12VAEncodeContext       *ctx = avctx->priv_data;
-    FFHWBaseEncodePicture *base_pic = &pic->base;
+    D3D12VAEncodeContext *ctx = avctx->priv_data;
+    D3D12VAEncodePicture *pic = base_pic->priv;
     uint64_t completion;
 
     av_assert0(base_pic->encode_issued);
@@ -186,12 +186,12 @@ static int d3d12va_encode_create_metadata_buffers(AVCodecContext *avctx,
 }
 
 static int d3d12va_encode_issue(AVCodecContext *avctx,
-                                const FFHWBaseEncodePicture *base_pic)
+                                FFHWBaseEncodePicture *base_pic)
 {
     FFHWBaseEncodeContext *base_ctx = avctx->priv_data;
     D3D12VAEncodeContext       *ctx = avctx->priv_data;
+    D3D12VAEncodePicture       *pic = base_pic->priv;
     AVD3D12VAFramesContext *frames_hwctx = base_ctx->input_frames->hwctx;
-    D3D12VAEncodePicture *pic = (D3D12VAEncodePicture *)base_pic;
     int err, i, j;
     HRESULT hr;
     char data[MAX_PARAM_BUFFER_SIZE];
@@ -288,7 +288,7 @@ static int d3d12va_encode_issue(AVCodecContext *avctx,
         goto fail;
 
     if (ctx->codec->init_picture_params) {
-        err = ctx->codec->init_picture_params(avctx, pic);
+        err = ctx->codec->init_picture_params(avctx, base_pic);
         if (err < 0) {
             av_log(avctx, AV_LOG_ERROR, "Failed to initialise picture "
                    "parameters: %d.\n", err);
@@ -333,9 +333,9 @@ static int d3d12va_encode_issue(AVCodecContext *avctx,
 
         i = 0;
         for (j = 0; j < base_pic->nb_refs[0]; j++)
-            d3d12_refs.ppTexture2Ds[i++] = ((D3D12VAEncodePicture *)base_pic->refs[0][j])->recon_surface->texture;
+            d3d12_refs.ppTexture2Ds[i++] = ((D3D12VAEncodePicture *)base_pic->refs[0][j]->priv)->recon_surface->texture;
         for (j = 0; j < base_pic->nb_refs[1]; j++)
-            d3d12_refs.ppTexture2Ds[i++] = ((D3D12VAEncodePicture *)base_pic->refs[1][j])->recon_surface->texture;
+            d3d12_refs.ppTexture2Ds[i++] = ((D3D12VAEncodePicture *)base_pic->refs[1][j]->priv)->recon_surface->texture;
     }
 
     input_args.PictureControlDesc.IntraRefreshFrameIndex  = 0;
@@ -515,10 +515,11 @@ fail:
 }
 
 static int d3d12va_encode_discard(AVCodecContext *avctx,
-                                  D3D12VAEncodePicture *pic)
+                                  FFHWBaseEncodePicture *base_pic)
 {
-    FFHWBaseEncodePicture *base_pic = &pic->base;
-    d3d12va_encode_wait(avctx, pic);
+    D3D12VAEncodePicture *pic = base_pic->priv;
+
+    d3d12va_encode_wait(avctx, base_pic);
 
     if (pic->output_buffer_ref) {
         av_log(avctx, AV_LOG_DEBUG, "Discard output for pic "
@@ -560,44 +561,33 @@ static int d3d12va_encode_free_rc_params(AVCodecContext *avctx)
     return 0;
 }
 
-static FFHWBaseEncodePicture *d3d12va_encode_alloc(AVCodecContext *avctx,
-                                                   const AVFrame *frame)
+static int d3d12va_encode_init(AVCodecContext *avctx, FFHWBaseEncodePicture *pic)
 {
     D3D12VAEncodeContext *ctx = avctx->priv_data;
-    D3D12VAEncodePicture *pic;
-
-    pic = av_mallocz(sizeof(*pic));
-    if (!pic)
-        return NULL;
+    D3D12VAEncodePicture *priv = pic->priv;
+    AVFrame *frame = pic->input_image;
 
     if (ctx->codec->picture_priv_data_size > 0) {
-        pic->base.priv_data = av_mallocz(ctx->codec->picture_priv_data_size);
-        if (!pic->base.priv_data) {
-            av_freep(&pic);
-            return NULL;
-        }
+        pic->codec_priv = av_mallocz(ctx->codec->picture_priv_data_size);
+        if (!pic->codec_priv)
+            return AVERROR(ENOMEM);
     }
 
-    pic->input_surface = (AVD3D12VAFrame *)frame->data[0];
+    priv->input_surface = (AVD3D12VAFrame *)frame->data[0];
 
-    return &pic->base;
+    return 0;
 }
 
-static int d3d12va_encode_free(AVCodecContext *avctx,
-                               FFHWBaseEncodePicture *base_pic)
+static int d3d12va_encode_free(AVCodecContext *avctx, FFHWBaseEncodePicture *pic)
 {
     D3D12VAEncodeContext *ctx = avctx->priv_data;
-    D3D12VAEncodePicture *pic = (D3D12VAEncodePicture *)base_pic;
+    D3D12VAEncodePicture *priv = pic->priv;
 
-    if (base_pic->encode_issued)
+    if (pic->encode_issued)
         d3d12va_encode_discard(avctx, pic);
 
     if (ctx->codec->free_picture_params)
-        ctx->codec->free_picture_params(pic);
-
-    ff_hw_base_encode_free(base_pic);
-
-    av_free(pic);
+        ctx->codec->free_picture_params(priv);
 
     return 0;
 }
@@ -680,14 +670,14 @@ end:
 }
 
 static int d3d12va_encode_output(AVCodecContext *avctx,
-                                 const FFHWBaseEncodePicture *base_pic, AVPacket *pkt)
+                                 FFHWBaseEncodePicture *base_pic, AVPacket *pkt)
 {
     FFHWBaseEncodeContext *base_ctx = avctx->priv_data;
-    D3D12VAEncodePicture *pic = (D3D12VAEncodePicture *)base_pic;
+    D3D12VAEncodePicture *pic = base_pic->priv;
     AVPacket *pkt_ptr = pkt;
     int err;
 
-    err = d3d12va_encode_wait(avctx, pic);
+    err = d3d12va_encode_wait(avctx, base_pic);
     if (err < 0)
         return err;
 
@@ -1389,7 +1379,9 @@ static int d3d12va_encode_create_recon_frames(AVCodecContext *avctx)
 }
 
 static const FFHWEncodePictureOperation d3d12va_type = {
-    .alloc  = &d3d12va_encode_alloc,
+    .priv_size = sizeof(D3D12VAEncodePicture),
+
+    .init   = &d3d12va_encode_init,
 
     .issue  = &d3d12va_encode_issue,
 
diff --git a/libavcodec/d3d12va_encode.h b/libavcodec/d3d12va_encode.h
index 51440428e4..3b0b8153d5 100644
--- a/libavcodec/d3d12va_encode.h
+++ b/libavcodec/d3d12va_encode.h
@@ -40,8 +40,6 @@ extern const AVCodecHWConfigInternal *const ff_d3d12va_encode_hw_configs[];
 #define D3D12VA_VIDEO_ENC_ASYNC_DEPTH 8
 
 typedef struct D3D12VAEncodePicture {
-    FFHWBaseEncodePicture base;
-
     int             header_size;
     int             aligned_header_size;
 
@@ -303,7 +301,7 @@ typedef struct D3D12VAEncodeType {
     int (*init_sequence_params)(AVCodecContext *avctx);
 
     int (*init_picture_params)(AVCodecContext *avctx,
-                               D3D12VAEncodePicture *pic);
+                               FFHWBaseEncodePicture *base_pic);
 
     void (*free_picture_params)(D3D12VAEncodePicture *pic);
 
diff --git a/libavcodec/d3d12va_encode_hevc.c b/libavcodec/d3d12va_encode_hevc.c
index 4a12ddd0dc..571ce57385 100644
--- a/libavcodec/d3d12va_encode_hevc.c
+++ b/libavcodec/d3d12va_encode_hevc.c
@@ -748,12 +748,12 @@ static void d3d12va_encode_hevc_free_picture_params(D3D12VAEncodePicture *pic)
 }
 
 static int d3d12va_encode_hevc_init_picture_params(AVCodecContext *avctx,
-                                                   D3D12VAEncodePicture *pic)
+                                                   FFHWBaseEncodePicture *base_pic)
 {
-    FFHWBaseEncodePicture                           *base_pic = &pic->base;
-    D3D12VAEncodeHEVCPicture                            *hpic = base_pic->priv_data;
+    D3D12VAEncodePicture                                 *pic = base_pic->priv;
+    D3D12VAEncodeHEVCPicture                            *hpic = base_pic->codec_priv;
     FFHWBaseEncodePicture                               *prev = base_pic->prev;
-    D3D12VAEncodeHEVCPicture                           *hprev = prev ? prev->priv_data : NULL;
+    D3D12VAEncodeHEVCPicture                           *hprev = prev ? prev->codec_priv : NULL;
     D3D12_VIDEO_ENCODER_REFERENCE_PICTURE_DESCRIPTOR_HEVC *pd = NULL;
     UINT                                           *ref_list0 = NULL, *ref_list1 = NULL;
     int i, idx = 0;
@@ -807,7 +807,7 @@ static int d3d12va_encode_hevc_init_picture_params(AVCodecContext *avctx,
             D3D12VAEncodeHEVCPicture *href;
 
             av_assert0(ref && ref->encode_order < base_pic->encode_order);
-            href = ref->priv_data;
+            href = ref->codec_priv;
 
             ref_list0[i] = idx;
             pd[idx].ReconstructedPictureResourceIndex = idx;
@@ -828,7 +828,7 @@ static int d3d12va_encode_hevc_init_picture_params(AVCodecContext *avctx,
             D3D12VAEncodeHEVCPicture *href;
 
             av_assert0(ref && ref->encode_order < base_pic->encode_order);
-            href = ref->priv_data;
+            href = ref->codec_priv;
 
             ref_list1[i] = idx;
             pd[idx].ReconstructedPictureResourceIndex = idx;
diff --git a/libavcodec/hw_base_encode.c b/libavcodec/hw_base_encode.c
index c1346e866c..35dac6e596 100644
--- a/libavcodec/hw_base_encode.c
+++ b/libavcodec/hw_base_encode.c
@@ -27,6 +27,19 @@
 #include "avcodec.h"
 #include "hw_base_encode.h"
 
+static int base_encode_pic_free(FFHWBaseEncodePicture *pic)
+{
+    av_frame_free(&pic->input_image);
+    av_frame_free(&pic->recon_image);
+
+    av_buffer_unref(&pic->opaque_ref);
+    av_freep(&pic->codec_priv);
+    av_freep(&pic->priv);
+    av_free(pic);
+
+    return 0;
+}
+
 static void hw_base_encode_add_ref(FFHWBaseEncodePicture *pic,
                                    FFHWBaseEncodePicture *target,
                                    int is_ref, int in_dpb, int prev)
@@ -370,6 +383,7 @@ static int hw_base_encode_clear_old(AVCodecContext *avctx, FFHWBaseEncodeContext
             else
                 ctx->pic_start = next;
             ctx->op->free(avctx, pic);
+            base_encode_pic_free(pic);
         } else {
             prev = pic;
         }
@@ -416,7 +430,7 @@ static int hw_base_encode_send_frame(AVCodecContext *avctx, FFHWBaseEncodeContex
         if (err < 0)
             return err;
 
-        pic = ctx->op->alloc(avctx, frame);
+        pic = av_mallocz(sizeof(*pic));
         if (!pic)
             return AVERROR(ENOMEM);
 
@@ -432,6 +446,12 @@ static int hw_base_encode_send_frame(AVCodecContext *avctx, FFHWBaseEncodeContex
             goto fail;
         }
 
+        pic->priv = av_mallocz(ctx->op->priv_size);
+        if (!pic->priv) {
+            err = AVERROR(ENOMEM);
+            goto fail;
+        }
+
         if (ctx->input_order == 0 || frame->pict_type == AV_PICTURE_TYPE_I)
             pic->force_idr = 1;
 
@@ -467,6 +487,9 @@ static int hw_base_encode_send_frame(AVCodecContext *avctx, FFHWBaseEncodeContex
             ctx->pic_end       = pic;
         }
 
+        err = ctx->op->init(avctx, pic);
+        if (err < 0)
+            goto fail;
     } else {
         ctx->end_of_stream = 1;
 
@@ -480,6 +503,7 @@ static int hw_base_encode_send_frame(AVCodecContext *avctx, FFHWBaseEncodeContex
 
 fail:
     ctx->op->free(avctx, pic);
+    base_encode_pic_free(pic);
     return err;
 }
 
@@ -529,7 +553,7 @@ int ff_hw_base_encode_receive_packet(FFHWBaseEncodeContext *ctx,
     AVFrame *frame = ctx->frame;
     int err;
 
-    av_assert0(ctx->op && ctx->op->alloc && ctx->op->issue &&
+    av_assert0(ctx->op && ctx->op->init && ctx->op->issue &&
                ctx->op->output && ctx->op->free);
 
 start:
@@ -737,17 +761,6 @@ fail:
     return err;
 }
 
-int ff_hw_base_encode_free(FFHWBaseEncodePicture *pic)
-{
-    av_frame_free(&pic->input_image);
-    av_frame_free(&pic->recon_image);
-
-    av_buffer_unref(&pic->opaque_ref);
-    av_freep(&pic->priv_data);
-
-    return 0;
-}
-
 int ff_hw_base_encode_init(AVCodecContext *avctx, FFHWBaseEncodeContext *ctx)
 {
     ctx->log_ctx = (void *)avctx;
diff --git a/libavcodec/hw_base_encode.h b/libavcodec/hw_base_encode.h
index ac0cc7b80c..e30b1e60ad 100644
--- a/libavcodec/hw_base_encode.h
+++ b/libavcodec/hw_base_encode.h
@@ -59,6 +59,11 @@ enum {
 };
 
 typedef struct FFHWBaseEncodePicture {
+    // API-specific private data
+    void *priv;
+    // Codec-specific private data
+    void *codec_priv;
+
     struct FFHWBaseEncodePicture *next;
 
     int64_t         display_order;
@@ -78,8 +83,6 @@ typedef struct FFHWBaseEncodePicture {
     AVFrame        *input_image;
     AVFrame        *recon_image;
 
-    void           *priv_data;
-
     // Whether this picture is a reference picture.
     int             is_reference;
 
@@ -104,15 +107,16 @@ typedef struct FFHWBaseEncodePicture {
 } FFHWBaseEncodePicture;
 
 typedef struct FFHWEncodePictureOperation {
-    // Alloc memory for the picture structure and initialize the API-specific internals
-    // based of the given frame.
-    FFHWBaseEncodePicture * (*alloc)(AVCodecContext *avctx, const AVFrame *frame);
+    // Size of API-specific internal picture data
+    size_t priv_size;
+    // Initialize API-specific internals
+    int (*init)(AVCodecContext *avctx, FFHWBaseEncodePicture *pic);
     // Issue the picture structure, which will send the frame surface to HW Encode API.
-    int (*issue)(AVCodecContext *avctx, const FFHWBaseEncodePicture *base_pic);
+    int (*issue)(AVCodecContext *avctx, FFHWBaseEncodePicture *pic);
     // Get the output AVPacket.
-    int (*output)(AVCodecContext *avctx, const FFHWBaseEncodePicture *base_pic, AVPacket *pkt);
+    int (*output)(AVCodecContext *avctx, FFHWBaseEncodePicture *pic, AVPacket *pkt);
     // Free the picture structure.
-    int (*free)(AVCodecContext *avctx, FFHWBaseEncodePicture *base_pic);
+    int (*free)(AVCodecContext *avctx, FFHWBaseEncodePicture *pic);
 }  FFHWEncodePictureOperation;
 
 typedef struct FFHWBaseEncodeContext {
@@ -228,8 +232,6 @@ int ff_hw_base_init_gop_structure(FFHWBaseEncodeContext *ctx, AVCodecContext *av
 int ff_hw_base_get_recon_format(FFHWBaseEncodeContext *ctx, const void *hwconfig,
                                 enum AVPixelFormat *fmt);
 
-int ff_hw_base_encode_free(FFHWBaseEncodePicture *pic);
-
 int ff_hw_base_encode_init(AVCodecContext *avctx, FFHWBaseEncodeContext *ctx);
 
 int ff_hw_base_encode_close(FFHWBaseEncodeContext *ctx);
diff --git a/libavcodec/vaapi_encode.c b/libavcodec/vaapi_encode.c
index afdf3fe3d8..b72e9035cb 100644
--- a/libavcodec/vaapi_encode.c
+++ b/libavcodec/vaapi_encode.c
@@ -135,14 +135,13 @@ static int vaapi_encode_make_misc_param_buffer(AVCodecContext *avctx,
                                           buffer, buffer_size);
 }
 
-static int vaapi_encode_wait(AVCodecContext *avctx,
-                             VAAPIEncodePicture *pic)
+static int vaapi_encode_wait(AVCodecContext *avctx, FFHWBaseEncodePicture *base_pic)
 {
 #if VA_CHECK_VERSION(1, 9, 0)
     FFHWBaseEncodeContext *base_ctx = avctx->priv_data;
 #endif
     VAAPIEncodeContext *ctx = avctx->priv_data;
-    FFHWBaseEncodePicture *base_pic = &pic->base;
+    VAAPIEncodePicture *pic = base_pic->priv;
     VAStatus vas;
 
     av_assert0(base_pic->encode_issued);
@@ -267,11 +266,11 @@ static int vaapi_encode_make_tile_slice(AVCodecContext *avctx,
 }
 
 static int vaapi_encode_issue(AVCodecContext *avctx,
-                              const FFHWBaseEncodePicture *base_pic)
+                              FFHWBaseEncodePicture *base_pic)
 {
     FFHWBaseEncodeContext *base_ctx = avctx->priv_data;
     VAAPIEncodeContext         *ctx = avctx->priv_data;
-    VAAPIEncodePicture *pic = (VAAPIEncodePicture*)base_pic;
+    VAAPIEncodePicture         *pic = base_pic->priv;
     VAAPIEncodeSlice *slice;
     VAStatus vas;
     int err, i;
@@ -364,7 +363,7 @@ static int vaapi_encode_issue(AVCodecContext *avctx,
     }
 
     if (ctx->codec->init_picture_params) {
-        err = ctx->codec->init_picture_params(avctx, pic);
+        err = ctx->codec->init_picture_params(avctx, base_pic);
         if (err < 0) {
             av_log(avctx, AV_LOG_ERROR, "Failed to initialise picture "
                    "parameters: %d.\n", err);
@@ -410,7 +409,7 @@ static int vaapi_encode_issue(AVCodecContext *avctx,
     if (ctx->va_packed_headers & VA_ENC_PACKED_HEADER_PICTURE &&
         ctx->codec->write_picture_header) {
         bit_len = 8 * sizeof(data);
-        err = ctx->codec->write_picture_header(avctx, pic, data, &bit_len);
+        err = ctx->codec->write_picture_header(avctx, base_pic, data, &bit_len);
         if (err < 0) {
             av_log(avctx, AV_LOG_ERROR, "Failed to write per-picture "
                    "header: %d.\n", err);
@@ -427,7 +426,7 @@ static int vaapi_encode_issue(AVCodecContext *avctx,
         for (i = 0;; i++) {
             size_t len = sizeof(data);
             int type;
-            err = ctx->codec->write_extra_buffer(avctx, pic, i, &type,
+            err = ctx->codec->write_extra_buffer(avctx, base_pic, i, &type,
                                                  data, &len);
             if (err == AVERROR_EOF)
                 break;
@@ -449,7 +448,7 @@ static int vaapi_encode_issue(AVCodecContext *avctx,
         for (i = 0;; i++) {
             int type;
             bit_len = 8 * sizeof(data);
-            err = ctx->codec->write_extra_header(avctx, pic, i, &type,
+            err = ctx->codec->write_extra_header(avctx, base_pic, i, &type,
                                                  data, &bit_len);
             if (err == AVERROR_EOF)
                 break;
@@ -493,7 +492,7 @@ static int vaapi_encode_issue(AVCodecContext *avctx,
         }
 
         if (ctx->codec->init_slice_params) {
-            err = ctx->codec->init_slice_params(avctx, pic, slice);
+            err = ctx->codec->init_slice_params(avctx, base_pic, slice);
             if (err < 0) {
                 av_log(avctx, AV_LOG_ERROR, "Failed to initialise slice "
                        "parameters: %d.\n", err);
@@ -773,15 +772,15 @@ end:
 }
 
 static int vaapi_encode_output(AVCodecContext *avctx,
-                               const FFHWBaseEncodePicture *base_pic, AVPacket *pkt)
+                               FFHWBaseEncodePicture *base_pic, AVPacket *pkt)
 {
     FFHWBaseEncodeContext *base_ctx = avctx->priv_data;
     VAAPIEncodeContext         *ctx = avctx->priv_data;
-    VAAPIEncodePicture         *pic = (VAAPIEncodePicture*)base_pic;
+    VAAPIEncodePicture         *pic = base_pic->priv;
     AVPacket *pkt_ptr = pkt;
     int err;
 
-    err = vaapi_encode_wait(avctx, pic);
+    err = vaapi_encode_wait(avctx, base_pic);
     if (err < 0)
         return err;
 
@@ -820,12 +819,11 @@ end:
     return err;
 }
 
-static int vaapi_encode_discard(AVCodecContext *avctx,
-                                VAAPIEncodePicture *pic)
+static int vaapi_encode_discard(AVCodecContext *avctx, FFHWBaseEncodePicture *base_pic)
 {
-    FFHWBaseEncodePicture *base_pic = &pic->base;
+    VAAPIEncodePicture *pic = base_pic->priv;
 
-    vaapi_encode_wait(avctx, pic);
+    vaapi_encode_wait(avctx, base_pic);
 
     if (pic->output_buffer_ref) {
         av_log(avctx, AV_LOG_DEBUG, "Discard output for pic "
@@ -839,56 +837,45 @@ static int vaapi_encode_discard(AVCodecContext *avctx,
     return 0;
 }
 
-static FFHWBaseEncodePicture *vaapi_encode_alloc(AVCodecContext *avctx,
-                                                 const AVFrame *frame)
+static int vaapi_encode_init(AVCodecContext *avctx, FFHWBaseEncodePicture *pic)
 {
     VAAPIEncodeContext *ctx = avctx->priv_data;
-    VAAPIEncodePicture *pic;
-
-    pic = av_mallocz(sizeof(*pic));
-    if (!pic)
-        return NULL;
+    VAAPIEncodePicture *priv = pic->priv;
+    AVFrame *frame = pic->input_image;
 
     if (ctx->codec->picture_priv_data_size > 0) {
-        pic->base.priv_data = av_mallocz(ctx->codec->picture_priv_data_size);
-        if (!pic->base.priv_data) {
-            av_freep(&pic);
-            return NULL;
-        }
+        pic->codec_priv = av_mallocz(ctx->codec->picture_priv_data_size);
+        if (!pic->codec_priv)
+            return AVERROR(ENOMEM);
     }
 
-    pic->input_surface = (VASurfaceID)(uintptr_t)frame->data[3];
-    pic->recon_surface = VA_INVALID_ID;
-    pic->output_buffer = VA_INVALID_ID;
+    priv->input_surface = (VASurfaceID)(uintptr_t)frame->data[3];
+    priv->recon_surface = VA_INVALID_ID;
+    priv->output_buffer = VA_INVALID_ID;
 
-    return &pic->base;
+    return 0;
 }
 
-static int vaapi_encode_free(AVCodecContext *avctx,
-                             FFHWBaseEncodePicture *base_pic)
+static int vaapi_encode_free(AVCodecContext *avctx, FFHWBaseEncodePicture *pic)
 {
-    VAAPIEncodePicture *pic = (VAAPIEncodePicture*)base_pic;
+    VAAPIEncodePicture *priv = pic->priv;
     int i;
 
-    if (base_pic->encode_issued)
+    if (pic->encode_issued)
         vaapi_encode_discard(avctx, pic);
 
-    if (pic->slices) {
-        for (i = 0; i < pic->nb_slices; i++)
-            av_freep(&pic->slices[i].codec_slice_params);
+    if (priv->slices) {
+        for (i = 0; i < priv->nb_slices; i++)
+            av_freep(&priv->slices[i].codec_slice_params);
     }
 
-    ff_hw_base_encode_free(base_pic);
-
-    av_freep(&pic->param_buffers);
-    av_freep(&pic->slices);
+    av_freep(&priv->param_buffers);
+    av_freep(&priv->slices);
     // Output buffer should already be destroyed.
-    av_assert0(pic->output_buffer == VA_INVALID_ID);
+    av_assert0(priv->output_buffer == VA_INVALID_ID);
 
-    av_freep(&pic->codec_picture_params);
-    av_freep(&pic->roi);
-
-    av_free(pic);
+    av_freep(&priv->codec_picture_params);
+    av_freep(&priv->roi);
 
     return 0;
 }
@@ -2090,7 +2077,9 @@ static av_cold int vaapi_encode_create_recon_frames(AVCodecContext *avctx)
 }
 
 static const FFHWEncodePictureOperation vaapi_op = {
-    .alloc  = &vaapi_encode_alloc,
+    .priv_size = sizeof(VAAPIEncodePicture),
+
+    .init = &vaapi_encode_init,
 
     .issue  = &vaapi_encode_issue,
 
diff --git a/libavcodec/vaapi_encode.h b/libavcodec/vaapi_encode.h
index d76cdb8662..c4f85397a2 100644
--- a/libavcodec/vaapi_encode.h
+++ b/libavcodec/vaapi_encode.h
@@ -63,8 +63,6 @@ typedef struct VAAPIEncodeSlice {
 } VAAPIEncodeSlice;
 
 typedef struct VAAPIEncodePicture {
-    FFHWBaseEncodePicture base;
-
 #if VA_CHECK_VERSION(1, 0, 0)
     // ROI regions.
     VAEncROI       *roi;
@@ -301,9 +299,9 @@ typedef struct VAAPIEncodeType {
     // Fill the parameter structures.
     int  (*init_sequence_params)(AVCodecContext *avctx);
     int   (*init_picture_params)(AVCodecContext *avctx,
-                                 VAAPIEncodePicture *pic);
+                                 FFHWBaseEncodePicture *pic);
     int     (*init_slice_params)(AVCodecContext *avctx,
-                                 VAAPIEncodePicture *pic,
+                                 FFHWBaseEncodePicture *pic,
                                  VAAPIEncodeSlice *slice);
 
     // The type used by the packed header: this should look like
@@ -318,7 +316,7 @@ typedef struct VAAPIEncodeType {
     int (*write_sequence_header)(AVCodecContext *avctx,
                                  char *data, size_t *data_len);
     int  (*write_picture_header)(AVCodecContext *avctx,
-                                 VAAPIEncodePicture *pic,
+                                 FFHWBaseEncodePicture *pic,
                                  char *data, size_t *data_len);
     int    (*write_slice_header)(AVCodecContext *avctx,
                                  VAAPIEncodePicture *pic,
@@ -330,7 +328,7 @@ typedef struct VAAPIEncodeType {
     // with increasing index argument until AVERROR_EOF is
     // returned.
     int    (*write_extra_buffer)(AVCodecContext *avctx,
-                                 VAAPIEncodePicture *pic,
+                                 FFHWBaseEncodePicture *pic,
                                  int index, int *type,
                                  char *data, size_t *data_len);
 
@@ -338,7 +336,7 @@ typedef struct VAAPIEncodeType {
     // with increasing index argument until AVERROR_EOF is
     // returned.
     int    (*write_extra_header)(AVCodecContext *avctx,
-                                 VAAPIEncodePicture *pic,
+                                 FFHWBaseEncodePicture *pic,
                                  int index, int *type,
                                  char *data, size_t *data_len);
 } VAAPIEncodeType;
diff --git a/libavcodec/vaapi_encode_av1.c b/libavcodec/vaapi_encode_av1.c
index dfdb424185..6d1be9fc07 100644
--- a/libavcodec/vaapi_encode_av1.c
+++ b/libavcodec/vaapi_encode_av1.c
@@ -466,12 +466,12 @@ end:
 }
 
 static int vaapi_encode_av1_init_picture_params(AVCodecContext *avctx,
-                                                VAAPIEncodePicture *vaapi_pic)
+                                                FFHWBaseEncodePicture *pic)
 {
     VAAPIEncodeContext              *ctx = avctx->priv_data;
     VAAPIEncodeAV1Context          *priv = avctx->priv_data;
-    const FFHWBaseEncodePicture     *pic = &vaapi_pic->base;
-    VAAPIEncodeAV1Picture          *hpic = pic->priv_data;
+    VAAPIEncodePicture        *vaapi_pic = pic->priv;
+    VAAPIEncodeAV1Picture          *hpic = pic->codec_priv;
     AV1RawOBU                    *fh_obu = &priv->fh;
     AV1RawFrameHeader                *fh = &fh_obu->obu.frame.header;
     VAEncPictureParameterBufferAV1 *vpic = vaapi_pic->codec_picture_params;
@@ -503,7 +503,7 @@ static int vaapi_encode_av1_init_picture_params(AVCodecContext *avctx,
         fh->frame_type = AV1_FRAME_INTER;
         fh->base_q_idx = priv->q_idx_p;
         ref = pic->refs[0][pic->nb_refs[0] - 1];
-        href = ref->priv_data;
+        href = ref->codec_priv;
         hpic->slot = !href->slot;
         hpic->last_idr_frame = href->last_idr_frame;
         fh->refresh_frame_flags = 1 << hpic->slot;
@@ -519,7 +519,7 @@ static int vaapi_encode_av1_init_picture_params(AVCodecContext *avctx,
         /** set the 2nd nearest frame in L0 as Golden frame. */
         if (pic->nb_refs[0] > 1) {
             ref = pic->refs[0][pic->nb_refs[0] - 2];
-            href = ref->priv_data;
+            href = ref->codec_priv;
             fh->ref_frame_idx[3] = href->slot;
             fh->ref_order_hint[href->slot] = ref->display_order - href->last_idr_frame;
             vpic->ref_frame_ctrl_l0.fields.search_idx1 = AV1_REF_FRAME_GOLDEN;
@@ -540,7 +540,7 @@ static int vaapi_encode_av1_init_picture_params(AVCodecContext *avctx,
         vpic->ref_frame_ctrl_l1.fields.search_idx0 = AV1_REF_FRAME_BWDREF;
 
         ref                            = pic->refs[0][pic->nb_refs[0] - 1];
-        href                           = ref->priv_data;
+        href                           = ref->codec_priv;
         hpic->last_idr_frame           = href->last_idr_frame;
         fh->primary_ref_frame          = href->slot;
         fh->ref_order_hint[href->slot] = ref->display_order - href->last_idr_frame;
@@ -549,7 +549,7 @@ static int vaapi_encode_av1_init_picture_params(AVCodecContext *avctx,
         }
 
         ref                            = pic->refs[1][pic->nb_refs[1] - 1];
-        href                           = ref->priv_data;
+        href                           = ref->codec_priv;
         fh->ref_order_hint[href->slot] = ref->display_order - href->last_idr_frame;
         for (i = AV1_REF_FRAME_GOLDEN; i < AV1_REFS_PER_FRAME; i++) {
             fh->ref_frame_idx[i] = href->slot;
@@ -634,7 +634,7 @@ static int vaapi_encode_av1_init_picture_params(AVCodecContext *avctx,
         for (int j = 0; j < pic->nb_refs[i]; j++) {
             FFHWBaseEncodePicture *ref_pic = pic->refs[i][j];
 
-            slot = ((VAAPIEncodeAV1Picture*)ref_pic->priv_data)->slot;
+            slot = ((VAAPIEncodeAV1Picture*)ref_pic->codec_priv)->slot;
             av_assert0(vpic->reference_frames[slot] == VA_INVALID_SURFACE);
 
             vpic->reference_frames[slot] = ((VAAPIEncodePicture *)ref_pic)->recon_surface;
@@ -732,7 +732,7 @@ end:
 }
 
 static int vaapi_encode_av1_init_slice_params(AVCodecContext *avctx,
-                                              VAAPIEncodePicture *pic,
+                                              FFHWBaseEncodePicture *base,
                                               VAAPIEncodeSlice *slice)
 {
     VAAPIEncodeAV1Context      *priv = avctx->priv_data;
@@ -754,7 +754,7 @@ static int vaapi_encode_av1_init_slice_params(AVCodecContext *avctx,
 }
 
 static int vaapi_encode_av1_write_picture_header(AVCodecContext *avctx,
-                                                 VAAPIEncodePicture *vaapi_pic,
+                                                 FFHWBaseEncodePicture *pic,
                                                  char *data, size_t *data_len)
 {
     VAAPIEncodeAV1Context     *priv = avctx->priv_data;
@@ -762,7 +762,7 @@ static int vaapi_encode_av1_write_picture_header(AVCodecContext *avctx,
     CodedBitstreamAV1Context *cbctx = priv->cbc->priv_data;
     AV1RawOBU               *fh_obu = &priv->fh;
     AV1RawFrameHeader       *rep_fh = &fh_obu->obu.frame_header;
-    const FFHWBaseEncodePicture *pic = &vaapi_pic->base;
+    VAAPIEncodePicture   *vaapi_pic = pic->priv;
     VAAPIEncodeAV1Picture *href;
     int ret = 0;
 
@@ -770,7 +770,7 @@ static int vaapi_encode_av1_write_picture_header(AVCodecContext *avctx,
     /** Pack repeat frame header. */
     if (pic->display_order > pic->encode_order) {
         memset(fh_obu, 0, sizeof(*fh_obu));
-        href = pic->refs[0][pic->nb_refs[0] - 1]->priv_data;
+        href = pic->refs[0][pic->nb_refs[0] - 1]->codec_priv;
         fh_obu->header.obu_type = AV1_OBU_FRAME_HEADER;
         fh_obu->header.obu_has_size_field = 1;
 
@@ -804,7 +804,7 @@ end:
 }
 
 static int vaapi_encode_av1_write_extra_header(AVCodecContext *avctx,
-                                               VAAPIEncodePicture *pic,
+                                               FFHWBaseEncodePicture *base_pic,
                                                int index, int *type,
                                                char *data, size_t *data_len)
 {
diff --git a/libavcodec/vaapi_encode_h264.c b/libavcodec/vaapi_encode_h264.c
index 84f306b95f..7dc4779dde 100644
--- a/libavcodec/vaapi_encode_h264.c
+++ b/libavcodec/vaapi_encode_h264.c
@@ -209,7 +209,7 @@ fail:
 }
 
 static int vaapi_encode_h264_write_extra_header(AVCodecContext *avctx,
-                                                VAAPIEncodePicture *pic,
+                                                FFHWBaseEncodePicture *base,
                                                 int index, int *type,
                                                 char *data, size_t *data_len)
 {
@@ -233,7 +233,7 @@ static int vaapi_encode_h264_write_extra_header(AVCodecContext *avctx,
                 goto fail;
         }
         if (priv->sei_needed & SEI_TIMING) {
-            if (pic->base.type == FF_HW_PICTURE_TYPE_IDR) {
+            if (base->type == FF_HW_PICTURE_TYPE_IDR) {
                 err = ff_cbs_sei_add_message(priv->cbc, au, 1,
                                              SEI_TYPE_BUFFERING_PERIOD,
                                              &priv->sei_buffering_period, NULL);
@@ -620,14 +620,14 @@ static int vaapi_encode_h264_init_sequence_params(AVCodecContext *avctx)
 }
 
 static int vaapi_encode_h264_init_picture_params(AVCodecContext *avctx,
-                                                 VAAPIEncodePicture *vaapi_pic)
+                                                 FFHWBaseEncodePicture *pic)
 {
     FFHWBaseEncodeContext       *base_ctx = avctx->priv_data;
     VAAPIEncodeH264Context          *priv = avctx->priv_data;
-    const FFHWBaseEncodePicture      *pic = &vaapi_pic->base;
-    VAAPIEncodeH264Picture          *hpic = pic->priv_data;
+    VAAPIEncodePicture         *vaapi_pic = pic->priv;
+    VAAPIEncodeH264Picture          *hpic = pic->codec_priv;
     FFHWBaseEncodePicture           *prev = pic->prev;
-    VAAPIEncodeH264Picture         *hprev = prev ? prev->priv_data : NULL;
+    VAAPIEncodeH264Picture         *hprev = prev ? prev->codec_priv : NULL;
     VAEncPictureParameterBufferH264 *vpic = vaapi_pic->codec_picture_params;
     int i, j = 0;
 
@@ -736,10 +736,10 @@ static int vaapi_encode_h264_init_picture_params(AVCodecContext *avctx,
             VAAPIEncodeH264Picture *href;
 
             av_assert0(ref && ref->encode_order < pic->encode_order);
-            href = ref->priv_data;
+            href = ref->codec_priv;
 
             vpic->ReferenceFrames[j++] = (VAPictureH264) {
-                .picture_id          = ((VAAPIEncodePicture *)ref)->recon_surface,
+                .picture_id          = ((VAAPIEncodePicture *)ref->priv)->recon_surface,
                 .frame_idx           = href->frame_num,
                 .flags               = VA_PICTURE_H264_SHORT_TERM_REFERENCE,
                 .TopFieldOrderCnt    = href->pic_order_cnt,
@@ -766,37 +766,36 @@ static int vaapi_encode_h264_init_picture_params(AVCodecContext *avctx,
 }
 
 static void vaapi_encode_h264_default_ref_pic_list(AVCodecContext *avctx,
-                                                   VAAPIEncodePicture *vaapi_pic,
-                                                   VAAPIEncodePicture **rpl0,
-                                                   VAAPIEncodePicture **rpl1,
+                                                   FFHWBaseEncodePicture *pic,
+                                                   FFHWBaseEncodePicture **rpl0,
+                                                   FFHWBaseEncodePicture **rpl1,
                                                    int *rpl_size)
 {
-    FFHWBaseEncodePicture *pic = &vaapi_pic->base;
     FFHWBaseEncodePicture *prev;
     VAAPIEncodeH264Picture *hp, *hn, *hc;
     int i, j, n = 0;
 
     prev = pic->prev;
     av_assert0(prev);
-    hp = pic->priv_data;
+    hp = pic->codec_priv;
 
     for (i = 0; i < pic->prev->nb_dpb_pics; i++) {
-        hn = prev->dpb[i]->priv_data;
+        hn = prev->dpb[i]->codec_priv;
         av_assert0(hn->frame_num < hp->frame_num);
 
         if (pic->type == FF_HW_PICTURE_TYPE_P) {
             for (j = n; j > 0; j--) {
-                hc = rpl0[j - 1]->base.priv_data;
+                hc = rpl0[j - 1]->codec_priv;
                 av_assert0(hc->frame_num != hn->frame_num);
                 if (hc->frame_num > hn->frame_num)
                     break;
                 rpl0[j] = rpl0[j - 1];
             }
-            rpl0[j] = (VAAPIEncodePicture *)prev->dpb[i];
+            rpl0[j] = prev->dpb[i];
 
         } else if (pic->type == FF_HW_PICTURE_TYPE_B) {
             for (j = n; j > 0; j--) {
-                hc = rpl0[j - 1]->base.priv_data;
+                hc = rpl0[j - 1]->codec_priv;
                 av_assert0(hc->pic_order_cnt != hp->pic_order_cnt);
                 if (hc->pic_order_cnt < hp->pic_order_cnt) {
                     if (hn->pic_order_cnt > hp->pic_order_cnt ||
@@ -808,10 +807,10 @@ static void vaapi_encode_h264_default_ref_pic_list(AVCodecContext *avctx,
                 }
                 rpl0[j] = rpl0[j - 1];
             }
-            rpl0[j] = (VAAPIEncodePicture *)prev->dpb[i];
+            rpl0[j] = prev->dpb[i];
 
             for (j = n; j > 0; j--) {
-                hc = rpl1[j - 1]->base.priv_data;
+                hc = rpl1[j - 1]->codec_priv;
                 av_assert0(hc->pic_order_cnt != hp->pic_order_cnt);
                 if (hc->pic_order_cnt > hp->pic_order_cnt) {
                     if (hn->pic_order_cnt < hp->pic_order_cnt ||
@@ -823,7 +822,7 @@ static void vaapi_encode_h264_default_ref_pic_list(AVCodecContext *avctx,
                 }
                 rpl1[j] = rpl1[j - 1];
             }
-            rpl1[j] = (VAAPIEncodePicture *)prev->dpb[i];
+            rpl1[j] = prev->dpb[i];
         }
 
         ++n;
@@ -835,7 +834,7 @@ static void vaapi_encode_h264_default_ref_pic_list(AVCodecContext *avctx,
                 break;
         }
         if (i == n)
-            FFSWAP(VAAPIEncodePicture*, rpl1[0], rpl1[1]);
+            FFSWAP(FFHWBaseEncodePicture *, rpl1[0], rpl1[1]);
     }
 
     if (pic->type == FF_HW_PICTURE_TYPE_P ||
@@ -843,7 +842,7 @@ static void vaapi_encode_h264_default_ref_pic_list(AVCodecContext *avctx,
         av_log(avctx, AV_LOG_DEBUG, "Default RefPicList0 for fn=%d/poc=%d:",
                hp->frame_num, hp->pic_order_cnt);
         for (i = 0; i < n; i++) {
-            hn = rpl0[i]->base.priv_data;
+            hn = rpl0[i]->codec_priv;
             av_log(avctx, AV_LOG_DEBUG, "  fn=%d/poc=%d",
                    hn->frame_num, hn->pic_order_cnt);
         }
@@ -853,7 +852,7 @@ static void vaapi_encode_h264_default_ref_pic_list(AVCodecContext *avctx,
         av_log(avctx, AV_LOG_DEBUG, "Default RefPicList1 for fn=%d/poc=%d:",
                hp->frame_num, hp->pic_order_cnt);
         for (i = 0; i < n; i++) {
-            hn = rpl1[i]->base.priv_data;
+            hn = rpl1[i]->codec_priv;
             av_log(avctx, AV_LOG_DEBUG, "  fn=%d/poc=%d",
                    hn->frame_num, hn->pic_order_cnt);
         }
@@ -864,12 +863,12 @@ static void vaapi_encode_h264_default_ref_pic_list(AVCodecContext *avctx,
 }
 
 static int vaapi_encode_h264_init_slice_params(AVCodecContext *avctx,
-                                               VAAPIEncodePicture *vaapi_pic,
+                                               FFHWBaseEncodePicture *pic,
                                                VAAPIEncodeSlice *slice)
 {
     VAAPIEncodeH264Context          *priv = avctx->priv_data;
-    const FFHWBaseEncodePicture      *pic = &vaapi_pic->base;
-    VAAPIEncodeH264Picture          *hpic = pic->priv_data;
+    VAAPIEncodePicture         *vaapi_pic = pic->priv;
+    VAAPIEncodeH264Picture          *hpic = pic->codec_priv;
     FFHWBaseEncodePicture           *prev = pic->prev;
     H264RawSPS                       *sps = &priv->raw_sps;
     H264RawPPS                       *pps = &priv->raw_pps;
@@ -931,7 +930,7 @@ static int vaapi_encode_h264_init_slice_params(AVCodecContext *avctx,
         } else {
             sh->adaptive_ref_pic_marking_mode_flag = 1;
             for (i = 0; i < discard; i++) {
-                VAAPIEncodeH264Picture *old = discard_list[i]->priv_data;
+                VAAPIEncodeH264Picture *old = discard_list[i]->codec_priv;
                 av_assert0(old->frame_num < hpic->frame_num);
                 sh->mmco[i].memory_management_control_operation = 1;
                 sh->mmco[i].difference_of_pic_nums_minus1 =
@@ -944,11 +943,11 @@ static int vaapi_encode_h264_init_slice_params(AVCodecContext *avctx,
     // If the intended references are not the first entries of RefPicListN
     // by default, use ref-pic-list-modification to move them there.
     if (pic->type == FF_HW_PICTURE_TYPE_P || pic->type == FF_HW_PICTURE_TYPE_B) {
-        VAAPIEncodePicture *def_l0[MAX_DPB_SIZE], *def_l1[MAX_DPB_SIZE];
+        FFHWBaseEncodePicture *def_l0[MAX_DPB_SIZE], *def_l1[MAX_DPB_SIZE];
         VAAPIEncodeH264Picture *href;
         int n;
 
-        vaapi_encode_h264_default_ref_pic_list(avctx, vaapi_pic,
+        vaapi_encode_h264_default_ref_pic_list(avctx, pic,
                                                def_l0, def_l1, &n);
 
         if (pic->type == FF_HW_PICTURE_TYPE_P) {
@@ -963,7 +962,7 @@ static int vaapi_encode_h264_init_slice_params(AVCodecContext *avctx,
             if (need_rplm) {
                 int pic_num = hpic->frame_num;
                 for (i = 0; i < pic->nb_refs[0]; i++) {
-                    href = pic->refs[0][i]->priv_data;
+                    href = pic->refs[0][i]->codec_priv;
                     av_assert0(href->frame_num != pic_num);
                     if (href->frame_num < pic_num) {
                         sh->rplm_l0[i].modification_of_pic_nums_idc = 0;
@@ -984,7 +983,7 @@ static int vaapi_encode_h264_init_slice_params(AVCodecContext *avctx,
             int n0 = 0, n1 = 0;
             for (i = 0; i < pic->nb_refs[0]; i++) {
                 av_assert0(pic->refs[0][i]);
-                href = pic->refs[0][i]->priv_data;
+                href = pic->refs[0][i]->codec_priv;
                 av_assert0(href->pic_order_cnt < hpic->pic_order_cnt);
                 if (pic->refs[0][i] != (FFHWBaseEncodePicture *)def_l0[n0])
                     need_rplm_l0 = 1;
@@ -993,7 +992,7 @@ static int vaapi_encode_h264_init_slice_params(AVCodecContext *avctx,
 
             for (i = 0; i < pic->nb_refs[1]; i++) {
                 av_assert0(pic->refs[1][i]);
-                href = pic->refs[1][i]->priv_data;
+                href = pic->refs[1][i]->codec_priv;
                 av_assert0(href->pic_order_cnt > hpic->pic_order_cnt);
                 if (pic->refs[1][i] != (FFHWBaseEncodePicture *)def_l1[n1])
                     need_rplm_l1 = 1;
@@ -1004,7 +1003,7 @@ static int vaapi_encode_h264_init_slice_params(AVCodecContext *avctx,
             if (need_rplm_l0) {
                 int pic_num = hpic->frame_num;
                 for (i = j = 0; i < pic->nb_refs[0]; i++) {
-                    href = pic->refs[0][i]->priv_data;
+                    href = pic->refs[0][i]->codec_priv;
                     av_assert0(href->frame_num != pic_num);
                     if (href->frame_num < pic_num) {
                         sh->rplm_l0[j].modification_of_pic_nums_idc = 0;
@@ -1026,7 +1025,7 @@ static int vaapi_encode_h264_init_slice_params(AVCodecContext *avctx,
             if (need_rplm_l1) {
                 int pic_num = hpic->frame_num;
                 for (i = j = 0; i < pic->nb_refs[1]; i++) {
-                    href = pic->refs[1][i]->priv_data;
+                    href = pic->refs[1][i]->codec_priv;
                     av_assert0(href->frame_num != pic_num);
                     if (href->frame_num < pic_num) {
                         sh->rplm_l1[j].modification_of_pic_nums_idc = 0;
diff --git a/libavcodec/vaapi_encode_h265.c b/libavcodec/vaapi_encode_h265.c
index 548029a645..6d59a98114 100644
--- a/libavcodec/vaapi_encode_h265.c
+++ b/libavcodec/vaapi_encode_h265.c
@@ -200,7 +200,7 @@ fail:
 }
 
 static int vaapi_encode_h265_write_extra_header(AVCodecContext *avctx,
-                                                VAAPIEncodePicture *pic,
+                                                FFHWBaseEncodePicture *base,
                                                 int index, int *type,
                                                 char *data, size_t *data_len)
 {
@@ -757,14 +757,14 @@ static int vaapi_encode_h265_init_sequence_params(AVCodecContext *avctx)
 }
 
 static int vaapi_encode_h265_init_picture_params(AVCodecContext *avctx,
-                                                 VAAPIEncodePicture *vaapi_pic)
+                                                 FFHWBaseEncodePicture *pic)
 {
     FFHWBaseEncodeContext       *base_ctx = avctx->priv_data;
     VAAPIEncodeH265Context          *priv = avctx->priv_data;
-    FFHWBaseEncodePicture            *pic = &vaapi_pic->base;
-    VAAPIEncodeH265Picture          *hpic = pic->priv_data;
+    VAAPIEncodePicture         *vaapi_pic = pic->priv;
+    VAAPIEncodeH265Picture          *hpic = pic->codec_priv;
     FFHWBaseEncodePicture           *prev = pic->prev;
-    VAAPIEncodeH265Picture         *hprev = prev ? prev->priv_data : NULL;
+    VAAPIEncodeH265Picture         *hprev = prev ? prev->codec_priv : NULL;
     VAEncPictureParameterBufferHEVC *vpic = vaapi_pic->codec_picture_params;
     int i, j = 0;
 
@@ -923,10 +923,10 @@ static int vaapi_encode_h265_init_picture_params(AVCodecContext *avctx,
             VAAPIEncodeH265Picture *href;
 
             av_assert0(ref && ref->encode_order < pic->encode_order);
-            href = ref->priv_data;
+            href = ref->codec_priv;
 
             vpic->reference_frames[j++] = (VAPictureHEVC) {
-                .picture_id    = ((VAAPIEncodePicture *)ref)->recon_surface,
+                .picture_id    = ((VAAPIEncodePicture *)ref->priv)->recon_surface,
                 .pic_order_cnt = href->pic_order_cnt,
                 .flags = (ref->display_order < pic->display_order ?
                           VA_PICTURE_HEVC_RPS_ST_CURR_BEFORE : 0) |
@@ -973,13 +973,13 @@ static int vaapi_encode_h265_init_picture_params(AVCodecContext *avctx,
 }
 
 static int vaapi_encode_h265_init_slice_params(AVCodecContext *avctx,
-                                               VAAPIEncodePicture *vaapi_pic,
+                                               FFHWBaseEncodePicture *pic,
                                                VAAPIEncodeSlice *slice)
 {
     FFHWBaseEncodeContext        *base_ctx = avctx->priv_data;
     VAAPIEncodeH265Context           *priv = avctx->priv_data;
-    const FFHWBaseEncodePicture       *pic = &vaapi_pic->base;
-    VAAPIEncodeH265Picture           *hpic = pic->priv_data;
+    VAAPIEncodePicture          *vaapi_pic = pic->priv;
+    VAAPIEncodeH265Picture           *hpic = pic->codec_priv;
     const H265RawSPS                  *sps = &priv->raw_sps;
     const H265RawPPS                  *pps = &priv->raw_pps;
     H265RawSliceHeader                 *sh = &priv->raw_slice.header;
@@ -1021,7 +1021,7 @@ static int vaapi_encode_h265_init_slice_params(AVCodecContext *avctx,
         rps_pics = 0;
         for (i = 0; i < MAX_REFERENCE_LIST_NUM; i++) {
             for (j = 0; j < pic->nb_refs[i]; j++) {
-                strp = pic->refs[i][j]->priv_data;
+                strp = pic->refs[i][j]->codec_priv;
                 rps_poc[rps_pics]  = strp->pic_order_cnt;
                 rps_used[rps_pics] = 1;
                 ++rps_pics;
@@ -1046,7 +1046,7 @@ static int vaapi_encode_h265_init_slice_params(AVCodecContext *avctx,
             if (j < pic->nb_refs[1])
                 continue;
 
-            strp = pic->dpb[i]->priv_data;
+            strp = pic->dpb[i]->codec_priv;
             rps_poc[rps_pics]  = strp->pic_order_cnt;
             rps_used[rps_pics] = 0;
             ++rps_pics;
diff --git a/libavcodec/vaapi_encode_mjpeg.c b/libavcodec/vaapi_encode_mjpeg.c
index 7667845c3d..5f8266e98a 100644
--- a/libavcodec/vaapi_encode_mjpeg.c
+++ b/libavcodec/vaapi_encode_mjpeg.c
@@ -147,7 +147,7 @@ fail:
 }
 
 static int vaapi_encode_mjpeg_write_extra_buffer(AVCodecContext *avctx,
-                                                 VAAPIEncodePicture *pic,
+                                                 FFHWBaseEncodePicture *base,
                                                  int index, int *type,
                                                  char *data, size_t *data_len)
 {
@@ -220,11 +220,11 @@ static int vaapi_encode_mjpeg_write_extra_buffer(AVCodecContext *avctx,
 }
 
 static int vaapi_encode_mjpeg_init_picture_params(AVCodecContext *avctx,
-                                                  VAAPIEncodePicture *vaapi_pic)
+                                                  FFHWBaseEncodePicture *pic)
 {
     FFHWBaseEncodeContext       *base_ctx = avctx->priv_data;
     VAAPIEncodeMJPEGContext         *priv = avctx->priv_data;
-    const FFHWBaseEncodePicture      *pic = &vaapi_pic->base;
+    VAAPIEncodePicture         *vaapi_pic = pic->priv;
     JPEGRawFrameHeader                *fh = &priv->frame_header;
     JPEGRawScanHeader                 *sh = &priv->scan.header;
     VAEncPictureParameterBufferJPEG *vpic = vaapi_pic->codec_picture_params;
@@ -414,7 +414,7 @@ static int vaapi_encode_mjpeg_init_picture_params(AVCodecContext *avctx,
 }
 
 static int vaapi_encode_mjpeg_init_slice_params(AVCodecContext *avctx,
-                                                VAAPIEncodePicture *pic,
+                                                FFHWBaseEncodePicture *base,
                                                 VAAPIEncodeSlice *slice)
 {
     VAAPIEncodeMJPEGContext         *priv = avctx->priv_data;
diff --git a/libavcodec/vaapi_encode_mpeg2.c b/libavcodec/vaapi_encode_mpeg2.c
index 30a333ecd4..5701e2e803 100644
--- a/libavcodec/vaapi_encode_mpeg2.c
+++ b/libavcodec/vaapi_encode_mpeg2.c
@@ -141,7 +141,7 @@ fail:
 }
 
 static int vaapi_encode_mpeg2_write_picture_header(AVCodecContext *avctx,
-                                                   VAAPIEncodePicture *pic,
+                                                   FFHWBaseEncodePicture *pic,
                                                    char *data, size_t *data_len)
 {
     VAAPIEncodeMPEG2Context *priv = avctx->priv_data;
@@ -418,10 +418,10 @@ static int vaapi_encode_mpeg2_init_sequence_params(AVCodecContext *avctx)
 }
 
 static int vaapi_encode_mpeg2_init_picture_params(AVCodecContext *avctx,
-                                                  VAAPIEncodePicture *vaapi_pic)
+                                                  FFHWBaseEncodePicture *pic)
 {
     VAAPIEncodeMPEG2Context          *priv = avctx->priv_data;
-    const FFHWBaseEncodePicture       *pic = &vaapi_pic->base;
+    VAAPIEncodePicture          *vaapi_pic = pic->priv;
     MPEG2RawPictureHeader              *ph = &priv->picture_header;
     MPEG2RawPictureCodingExtension    *pce = &priv->picture_coding_extension.data.picture_coding;
     VAEncPictureParameterBufferMPEG2 *vpic = vaapi_pic->codec_picture_params;
@@ -460,12 +460,12 @@ static int vaapi_encode_mpeg2_init_picture_params(AVCodecContext *avctx,
         break;
     case FF_HW_PICTURE_TYPE_P:
         vpic->picture_type = VAEncPictureTypePredictive;
-        vpic->forward_reference_picture = ((VAAPIEncodePicture *)pic->refs[0][0])->recon_surface;
+        vpic->forward_reference_picture = ((VAAPIEncodePicture *)pic->refs[0][0]->priv)->recon_surface;
         break;
     case FF_HW_PICTURE_TYPE_B:
         vpic->picture_type = VAEncPictureTypeBidirectional;
-        vpic->forward_reference_picture  = ((VAAPIEncodePicture *)pic->refs[0][0])->recon_surface;
-        vpic->backward_reference_picture = ((VAAPIEncodePicture *)pic->refs[1][0])->recon_surface;
+        vpic->forward_reference_picture  = ((VAAPIEncodePicture *)pic->refs[0][0]->priv)->recon_surface;
+        vpic->backward_reference_picture = ((VAAPIEncodePicture *)pic->refs[1][0]->priv)->recon_surface;
         break;
     default:
         av_assert0(0 && "invalid picture type");
@@ -481,10 +481,9 @@ static int vaapi_encode_mpeg2_init_picture_params(AVCodecContext *avctx,
 }
 
 static int vaapi_encode_mpeg2_init_slice_params(AVCodecContext *avctx,
-                                                VAAPIEncodePicture *vaapi_pic,
+                                                FFHWBaseEncodePicture *pic,
                                                 VAAPIEncodeSlice *slice)
 {
-    const FFHWBaseEncodePicture       *pic = &vaapi_pic->base;
     VAAPIEncodeMPEG2Context          *priv = avctx->priv_data;
     VAEncSliceParameterBufferMPEG2 *vslice = slice->codec_slice_params;
     int qp;
diff --git a/libavcodec/vaapi_encode_vp8.c b/libavcodec/vaapi_encode_vp8.c
index b2576f4a45..ad88af63eb 100644
--- a/libavcodec/vaapi_encode_vp8.c
+++ b/libavcodec/vaapi_encode_vp8.c
@@ -74,9 +74,9 @@ static int vaapi_encode_vp8_init_sequence_params(AVCodecContext *avctx)
 }
 
 static int vaapi_encode_vp8_init_picture_params(AVCodecContext *avctx,
-                                                VAAPIEncodePicture *vaapi_pic)
+                                                FFHWBaseEncodePicture *pic)
 {
-    const FFHWBaseEncodePicture     *pic = &vaapi_pic->base;
+    VAAPIEncodePicture        *vaapi_pic = pic->priv;
     VAAPIEncodeVP8Context          *priv = avctx->priv_data;
     VAEncPictureParameterBufferVP8 *vpic = vaapi_pic->codec_picture_params;
     int i;
@@ -103,7 +103,7 @@ static int vaapi_encode_vp8_init_picture_params(AVCodecContext *avctx,
         vpic->ref_last_frame =
         vpic->ref_gf_frame   =
         vpic->ref_arf_frame  =
-            ((VAAPIEncodePicture *)pic->refs[0][0])->recon_surface;
+            ((VAAPIEncodePicture *)pic->refs[0][0]->priv)->recon_surface;
         break;
     default:
         av_assert0(0 && "invalid picture type");
@@ -129,7 +129,7 @@ static int vaapi_encode_vp8_init_picture_params(AVCodecContext *avctx,
 }
 
 static int vaapi_encode_vp8_write_quant_table(AVCodecContext *avctx,
-                                              VAAPIEncodePicture *pic,
+                                              FFHWBaseEncodePicture *base_pic,
                                               int index, int *type,
                                               char *data, size_t *data_len)
 {
@@ -147,7 +147,7 @@ static int vaapi_encode_vp8_write_quant_table(AVCodecContext *avctx,
 
     memset(&quant, 0, sizeof(quant));
 
-    if (pic->base.type == FF_HW_PICTURE_TYPE_P)
+    if (base_pic->type == FF_HW_PICTURE_TYPE_P)
         q = priv->q_index_p;
     else
         q = priv->q_index_i;
diff --git a/libavcodec/vaapi_encode_vp9.c b/libavcodec/vaapi_encode_vp9.c
index 7ce24cb864..ab925e24b5 100644
--- a/libavcodec/vaapi_encode_vp9.c
+++ b/libavcodec/vaapi_encode_vp9.c
@@ -77,12 +77,12 @@ static int vaapi_encode_vp9_init_sequence_params(AVCodecContext *avctx)
 }
 
 static int vaapi_encode_vp9_init_picture_params(AVCodecContext *avctx,
-                                                VAAPIEncodePicture *vaapi_pic)
+                                                FFHWBaseEncodePicture *pic)
 {
     FFHWBaseEncodeContext      *base_ctx = avctx->priv_data;
     VAAPIEncodeVP9Context          *priv = avctx->priv_data;
-    const FFHWBaseEncodePicture     *pic = &vaapi_pic->base;
-    VAAPIEncodeVP9Picture          *hpic = pic->priv_data;
+    VAAPIEncodePicture        *vaapi_pic = pic->priv;
+    VAAPIEncodeVP9Picture          *hpic = pic->codec_priv;
     VAEncPictureParameterBufferVP9 *vpic = vaapi_pic->codec_picture_params;
     int i;
     int num_tile_columns;
@@ -106,7 +106,7 @@ static int vaapi_encode_vp9_init_picture_params(AVCodecContext *avctx,
     case FF_HW_PICTURE_TYPE_P:
         av_assert0(!pic->nb_refs[1]);
         {
-            VAAPIEncodeVP9Picture *href = pic->refs[0][0]->priv_data;
+            VAAPIEncodeVP9Picture *href = pic->refs[0][0]->codec_priv;
             av_assert0(href->slot == 0 || href->slot == 1);
 
             if (base_ctx->max_b_depth > 0) {
@@ -124,8 +124,8 @@ static int vaapi_encode_vp9_init_picture_params(AVCodecContext *avctx,
     case FF_HW_PICTURE_TYPE_B:
         av_assert0(pic->nb_refs[0] && pic->nb_refs[1]);
         {
-            VAAPIEncodeVP9Picture *href0 = pic->refs[0][0]->priv_data,
-                                  *href1 = pic->refs[1][0]->priv_data;
+            VAAPIEncodeVP9Picture *href0 = pic->refs[0][0]->codec_priv,
+                                  *href1 = pic->refs[1][0]->codec_priv;
             av_assert0(href0->slot < pic->b_depth + 1 &&
                        href1->slot < pic->b_depth + 1);
 
@@ -163,9 +163,9 @@ static int vaapi_encode_vp9_init_picture_params(AVCodecContext *avctx,
         for (int j = 0; j < pic->nb_refs[i]; j++) {
             FFHWBaseEncodePicture *ref_pic = pic->refs[i][j];
             int slot;
-            slot = ((VAAPIEncodeVP9Picture*)ref_pic->priv_data)->slot;
+            slot = ((VAAPIEncodeVP9Picture*)ref_pic->codec_priv)->slot;
             av_assert0(vpic->reference_frames[slot] == VA_INVALID_SURFACE);
-            vpic->reference_frames[slot] = ((VAAPIEncodePicture *)ref_pic)->recon_surface;
+            vpic->reference_frames[slot] = ((VAAPIEncodePicture *)ref_pic->priv)->recon_surface;
         }
     }