FFmpeg-基础组件-AVFrame

本章主要介绍FFmpeg基础组件AVFrame.

1.结构体成员

我们把所有的代码先粘贴上来，在后边一个一个解释。

typedef struct AVFrame {
#define AV_NUM_DATA_POINTERS 8
    /**
     * pointer to the picture/channel planes.
     * This might be different from the first allocated byte. For video,
     * it could even point to the end of the image data.
     *
     * All pointers in data and extended_data must point into one of the
     * AVBufferRef in buf or extended_buf.
     *
     * Some decoders access areas outside 0,0 - width,height, please
     * see avcodec_align_dimensions2(). Some filters and swscale can read
     * up to 16 bytes beyond the planes, if these filters are to be used,
     * then 16 extra bytes must be allocated.
     *
     * NOTE: Pointers not needed by the format MUST be set to NULL.
     *
     * @attention In case of video, the data[] pointers can point to the
     * end of image data in order to reverse line order, when used in
     * combination with negative values in the linesize[] array.
     */
    uint8_t *data[AV_NUM_DATA_POINTERS];

    /**
     * For video, a positive or negative value, which is typically indicating
     * the size in bytes of each picture line, but it can also be:
     * - the negative byte size of lines for vertical flipping
     *   (with data[n] pointing to the end of the data
     * - a positive or negative multiple of the byte size as for accessing
     *   even and odd fields of a frame (possibly flipped)
     *
     * For audio, only linesize[0] may be set. For planar audio, each channel
     * plane must be the same size.
     *
     * For video the linesizes should be multiples of the CPUs alignment
     * preference, this is 16 or 32 for modern desktop CPUs.
     * Some code requires such alignment other code can be slower without
     * correct alignment, for yet other it makes no difference.
     *
     * @note The linesize may be larger than the size of usable data -- there
     * may be extra padding present for performance reasons.
     *
     * @attention In case of video, line size values can be negative to achieve
     * a vertically inverted iteration over image lines.
     */
    int linesize[AV_NUM_DATA_POINTERS];

    /**
     * pointers to the data planes/channels.
     *
     * For video, this should simply point to data[].
     *
     * For planar audio, each channel has a separate data pointer, and
     * linesize[0] contains the size of each channel buffer.
     * For packed audio, there is just one data pointer, and linesize[0]
     * contains the total size of the buffer for all channels.
     *
     * Note: Both data and extended_data should always be set in a valid frame,
     * but for planar audio with more channels that can fit in data,
     * extended_data must be used in order to access all channels.
     */
    uint8_t **extended_data;

    /**
     * @name Video dimensions
     * Video frames only. The coded dimensions (in pixels) of the video frame,
     * i.e. the size of the rectangle that contains some well-defined values.
     *
     * @note The part of the frame intended for display/presentation is further
     * restricted by the @ref cropping "Cropping rectangle".
     * @{
     */
    int width, height;
    /**
     * @}
     */

    /**
     * number of audio samples (per channel) described by this frame
     */
    int nb_samples;

    /**
     * format of the frame, -1 if unknown or unset
     * Values correspond to enum AVPixelFormat for video frames,
     * enum AVSampleFormat for audio)
     */
    int format;

    /**
     * 1 -> keyframe, 0-> not
     */
    int key_frame;

    /**
     * Picture type of the frame.
     */
    enum AVPictureType pict_type;

    /**
     * Sample aspect ratio for the video frame, 0/1 if unknown/unspecified.
     */
    AVRational sample_aspect_ratio;

    /**
     * Presentation timestamp in time_base units (time when frame should be shown to user).
     */
    int64_t pts;

    /**
     * DTS copied from the AVPacket that triggered returning this frame. (if frame threading isn't used)
     * This is also the Presentation time of this AVFrame calculated from
     * only AVPacket.dts values without pts values.
     */
    int64_t pkt_dts;

    /**
     * Time base for the timestamps in this frame.
     * In the future, this field may be set on frames output by decoders or
     * filters, but its value will be by default ignored on input to encoders
     * or filters.
     */
    AVRational time_base;

    /**
     * picture number in bitstream order
     */
    int coded_picture_number;
    /**
     * picture number in display order
     */
    int display_picture_number;

    /**
     * quality (between 1 (good) and FF_LAMBDA_MAX (bad))
     */
    int quality;

    /**
     * for some private data of the user
     */
    void *opaque;

    /**
     * When decoding, this signals how much the picture must be delayed.
     * extra_delay = repeat_pict / (2*fps)
     */
    int repeat_pict;

    /**
     * The content of the picture is interlaced.
     */
    int interlaced_frame;

    /**
     * If the content is interlaced, is top field displayed first.
     */
    int top_field_first;

    /**
     * Tell user application that palette has changed from previous frame.
     */
    int palette_has_changed;

    /**
     * reordered opaque 64 bits (generally an integer or a double precision float
     * PTS but can be anything).
     * The user sets AVCodecContext.reordered_opaque to represent the input at
     * that time,
     * the decoder reorders values as needed and sets AVFrame.reordered_opaque
     * to exactly one of the values provided by the user through AVCodecContext.reordered_opaque
     */
    int64_t reordered_opaque;

    /**
     * Sample rate of the audio data.
     */
    int sample_rate;

    /**
     * Channel layout of the audio data.
     */
    uint64_t channel_layout;

    /**
     * AVBuffer references backing the data for this frame. All the pointers in
     * data and extended_data must point inside one of the buffers in buf or
     * extended_buf. This array must be filled contiguously -- if buf[i] is
     * non-NULL then buf[j] must also be non-NULL for all j < i.
     *
     * There may be at most one AVBuffer per data plane, so for video this array
     * always contains all the references. For planar audio with more than
     * AV_NUM_DATA_POINTERS channels, there may be more buffers than can fit in
     * this array. Then the extra AVBufferRef pointers are stored in the
     * extended_buf array.
     */
    AVBufferRef *buf[AV_NUM_DATA_POINTERS];

    /**
     * For planar audio which requires more than AV_NUM_DATA_POINTERS
     * AVBufferRef pointers, this array will hold all the references which
     * cannot fit into AVFrame.buf.
     *
     * Note that this is different from AVFrame.extended_data, which always
     * contains all the pointers. This array only contains the extra pointers,
     * which cannot fit into AVFrame.buf.
     *
     * This array is always allocated using av_malloc() by whoever constructs
     * the frame. It is freed in av_frame_unref().
     */
    AVBufferRef **extended_buf;
    /**
     * Number of elements in extended_buf.
     */
    int        nb_extended_buf;

    AVFrameSideData **side_data;
    int            nb_side_data;

/**
 * @defgroup lavu_frame_flags AV_FRAME_FLAGS
 * @ingroup lavu_frame
 * Flags describing additional frame properties.
 *
 * @{
 */

/**
 * The frame data may be corrupted, e.g. due to decoding errors.
 */
#define AV_FRAME_FLAG_CORRUPT       (1 << 0)
/**
 * A flag to mark the frames which need to be decoded, but shouldn't be output.
 */
#define AV_FRAME_FLAG_DISCARD   (1 << 2)
/**
 * @}
 */

    /**
     * Frame flags, a combination of @ref lavu_frame_flags
     */
    int flags;

    /**
     * MPEG vs JPEG YUV range.
     * - encoding: Set by user
     * - decoding: Set by libavcodec
     */
    enum AVColorRange color_range;

    enum AVColorPrimaries color_primaries;

    enum AVColorTransferCharacteristic color_trc;

    /**
     * YUV colorspace type.
     * - encoding: Set by user
     * - decoding: Set by libavcodec
     */
    enum AVColorSpace colorspace;

    enum AVChromaLocation chroma_location;

    /**
     * frame timestamp estimated using various heuristics, in stream time base
     * - encoding: unused
     * - decoding: set by libavcodec, read by user.
     */
    int64_t best_effort_timestamp;

    /**
     * reordered pos from the last AVPacket that has been input into the decoder
     * - encoding: unused
     * - decoding: Read by user.
     */
    int64_t pkt_pos;

    /**
     * duration of the corresponding packet, expressed in
     * AVStream->time_base units, 0 if unknown.
     * - encoding: unused
     * - decoding: Read by user.
     */
    int64_t pkt_duration;

    /**
     * metadata.
     * - encoding: Set by user.
     * - decoding: Set by libavcodec.
     */
    AVDictionary *metadata;

    /**
     * decode error flags of the frame, set to a combination of
     * FF_DECODE_ERROR_xxx flags if the decoder produced a frame, but there
     * were errors during the decoding.
     * - encoding: unused
     * - decoding: set by libavcodec, read by user.
     */
    int decode_error_flags;
#define FF_DECODE_ERROR_INVALID_BITSTREAM   1
#define FF_DECODE_ERROR_MISSING_REFERENCE   2
#define FF_DECODE_ERROR_CONCEALMENT_ACTIVE  4
#define FF_DECODE_ERROR_DECODE_SLICES       8

    /**
     * number of audio channels, only used for audio.
     * - encoding: unused
     * - decoding: Read by user.
     */
    int channels;

    /**
     * size of the corresponding packet containing the compressed
     * frame.
     * It is set to a negative value if unknown.
     * - encoding: unused
     * - decoding: set by libavcodec, read by user.
     */
    int pkt_size;

    /**
     * For hwaccel-format frames, this should be a reference to the
     * AVHWFramesContext describing the frame.
     */
    AVBufferRef *hw_frames_ctx;

    /**
     * AVBufferRef for free use by the API user. FFmpeg will never check the
     * contents of the buffer ref. FFmpeg calls av_buffer_unref() on it when
     * the frame is unreferenced. av_frame_copy_props() calls create a new
     * reference with av_buffer_ref() for the target frame's opaque_ref field.
     *
     * This is unrelated to the opaque field, although it serves a similar
     * purpose.
     */
    AVBufferRef *opaque_ref;

    /**
     * @anchor cropping
     * @name Cropping
     * Video frames only. The number of pixels to discard from the the
     * top/bottom/left/right border of the frame to obtain the sub-rectangle of
     * the frame intended for presentation.
     * @{
     */
    size_t crop_top;
    size_t crop_bottom;
    size_t crop_left;
    size_t crop_right;
    /**
     * @}
     */

    /**
     * AVBufferRef for internal use by a single libav* library.
     * Must not be used to transfer data between libraries.
     * Has to be NULL when ownership of the frame leaves the respective library.
     *
     * Code outside the FFmpeg libs should never check or change the contents of the buffer ref.
     *
     * FFmpeg calls av_buffer_unref() on it when the frame is unreferenced.
     * av_frame_copy_props() calls create a new reference with av_buffer_ref()
     * for the target frame's private_ref field.
     */
    AVBufferRef *private_ref;
} AVFrame;

AVFrame中核心成员，我们常用的就是下面几个

typedef struct AVFrame {
...
	uint8_t *data[AV_NUM_DATA_POINTERS]
	int linesize[AV_NUM_DATA_POINTERS];
	uint8_t **extended_data;uint8_t **extended_data;
	int width, height;
	int format;
	int key_frame;
	int64_t pts;
	int64_t pkt_dts;
	AVRational time_base;
	AVBufferRef *buf[AV_NUM_DATA_POINTERS];
...
}

这个结构体主要保存解码后的YUV数据，特别注意的是AVBufferRef *buf[AV_NUM_DATA_POINTERS];这个成员，因为它用来进行分配内存的释放。
其实在很多地方用到了AVBufferRef ，原理就是其中内部使用了一个引用，当引用为1的时候，就把内部指向的memory释放掉。

/**
 * A reference counted buffer type. It is opaque and is meant to be used through
 * references (AVBufferRef).
 */
typedef struct AVBuffer AVBuffer;

/**
 * A reference to a data buffer.
 *
 * The size of this struct is not a part of the public ABI and it is not meant
 * to be allocated directly.
 */
typedef struct AVBufferRef {
    AVBuffer *buffer;

    /**
     * The data buffer. It is considered writable if and only if
     * this is the only reference to the buffer, in which case
     * av_buffer_is_writable() returns 1.
     */
    uint8_t *data;
    /**
     * Size of data in bytes.
     */
    size_t   size;
} AVBufferRef;

其中data就是yuv数据，size就是数据大小，我们看到里面的AVBuffer ，其详细结构如下，其中refcount就是引用数量，free就是释放函数

struct AVBuffer {
    uint8_t *data; /**< data described by this buffer */
    size_t size; /**< size of data in bytes */

    /**
     *  number of existing AVBufferRef instances referring to this buffer
     */
    atomic_uint refcount;

    /**
     * a callback for freeing the data
     */
    void (*free)(void *opaque, uint8_t *data);

    /**
     * an opaque pointer, to be used by the freeing callback
     */
    void *opaque;

    /**
     * A combination of AV_BUFFER_FLAG_*
     */
    int flags;

    /**
     * A combination of BUFFER_FLAG_*
     */
    int flags_internal;
};

关于这一块智能引用，其提供了一些函数族

/**
 * Allocate an AVBuffer of the given size using av_malloc().
 *
 * @return an AVBufferRef of given size or NULL when out of memory
 */
AVBufferRef *av_buffer_alloc(size_t size);

/**
 * Same as av_buffer_alloc(), except the returned buffer will be initialized
 * to zero.
 */
AVBufferRef *av_buffer_allocz(size_t size);

/**
 * Create an AVBuffer from an existing array.
 *
 * If this function is successful, data is owned by the AVBuffer. The caller may
 * only access data through the returned AVBufferRef and references derived from
 * it.
 * If this function fails, data is left untouched.
 * @param data   data array
 * @param size   size of data in bytes
 * @param free   a callback for freeing this buffer's data
 * @param opaque parameter to be got for processing or passed to free
 * @param flags  a combination of AV_BUFFER_FLAG_*
 *
 * @return an AVBufferRef referring to data on success, NULL on failure.
 */
AVBufferRef *av_buffer_create(uint8_t *data, size_t size,
                              void (*free)(void *opaque, uint8_t *data),
                              void *opaque, int flags);

/**
 * Default free callback, which calls av_free() on the buffer data.
 * This function is meant to be passed to av_buffer_create(), not called
 * directly.
 */
void av_buffer_default_free(void *opaque, uint8_t *data);

/**
 * Create a new reference to an AVBuffer.
 *
 * @return a new AVBufferRef referring to the same AVBuffer as buf or NULL on
 * failure.
 */
AVBufferRef *av_buffer_ref(const AVBufferRef *buf);

/**
 * Free a given reference and automatically free the buffer if there are no more
 * references to it.
 *
 * @param buf the reference to be freed. The pointer is set to NULL on return.
 */
void av_buffer_unref(AVBufferRef **buf);

/**
 * @return 1 if the caller may write to the data referred to by buf (which is
 * true if and only if buf is the only reference to the underlying AVBuffer).
 * Return 0 otherwise.
 * A positive answer is valid until av_buffer_ref() is called on buf.
 */
int av_buffer_is_writable(const AVBufferRef *buf);

/**
 * @return the opaque parameter set by av_buffer_create.
 */
void *av_buffer_get_opaque(const AVBufferRef *buf);

int av_buffer_get_ref_count(const AVBufferRef *buf);

这些函数族用户是用不到的，一般在ffmpeg内部模块中使用到。比如AVFrame函数族中。

其余根据字面意思很好理解。其中linesize表示每个plane的步长。
比如YUV420的数，其数据长度大小为：

  if (frame->pixel_format == TOPSCODEC_PIX_FMT_I420 ||
      frame->pixel_format == TOPSCODEC_PIX_FMT_NV12 ||
      frame->pixel_format == TOPSCODEC_PIX_FMT_NV21 ||
      frame->pixel_format == TOPSCODEC_PIX_FMT_P010 ||
      frame->pixel_format == TOPSCODEC_PIX_FMT_P010LE) {
    int Y = frame->linesize[0] * frame->height;
    int U = frame->linesize[1] * ((frame->height + 1) / 2);
    int V = frame->linesize[2] * ((frame->height + 1) / 2);
    }

也就是说yuv数据的宽度是多大。
其实这个并不用手工来计算，ffmpeg框架提供了专用的函数来处理：

int av_image_fill_plane_sizes(size_t sizes[4], enum AVPixelFormat pix_fmt,
                              int height, const ptrdiff_t linesizes[4])

这个函数用来计算每个plane的长度是多少，通过传入参数我们很容易看到height.原理和我们上面一样。

下面这个函数比较特殊，可以计算linesizes，也就是每个通道的stride.

int av_image_fill_linesizes(int linesizes[4], enum AVPixelFormat pix_fmt, int width)

这个函数专门通过width和pix_fmt来计算linesizes[4]，这样就不用我们通过手动来去计算了。

还有另外一个函数用来计算每个planes数据长度

该类函数族ffmpeg提供了很多，具体在imgutils.h中

/**
 * Compute the size of an image line with format pix_fmt and width
 * width for the plane plane.
 *
 * @return the computed size in bytes
 */
int av_image_get_linesize(enum AVPixelFormat pix_fmt, int width, int plane);

/**
 * Fill plane linesizes for an image with pixel format pix_fmt and
 * width width.
 *
 * @param linesizes array to be filled with the linesize for each plane
 * @return >= 0 in case of success, a negative error code otherwise
 */
int av_image_fill_linesizes(int linesizes[4], enum AVPixelFormat pix_fmt, int width);

/**
 * Fill plane sizes for an image with pixel format pix_fmt and height height.
 *
 * @param size the array to be filled with the size of each image plane
 * @param linesizes the array containing the linesize for each
 *        plane, should be filled by av_image_fill_linesizes()
 * @return >= 0 in case of success, a negative error code otherwise
 *
 * @note The linesize parameters have the type ptrdiff_t here, while they are
 *       int for av_image_fill_linesizes().
 */
int av_image_fill_plane_sizes(size_t size[4], enum AVPixelFormat pix_fmt,
                              int height, const ptrdiff_t linesizes[4]);

/**
 * Fill plane data pointers for an image with pixel format pix_fmt and
 * height height.
 *
 * @param data pointers array to be filled with the pointer for each image plane
 * @param ptr the pointer to a buffer which will contain the image
 * @param linesizes the array containing the linesize for each
 * plane, should be filled by av_image_fill_linesizes()
 * @return the size in bytes required for the image buffer, a negative
 * error code in case of failure
 */
int av_image_fill_pointers(uint8_t *data[4], enum AVPixelFormat pix_fmt, int height,
                           uint8_t *ptr, const int linesizes[4]);

/**
 * Allocate an image with size w and h and pixel format pix_fmt, and
 * fill pointers and linesizes accordingly.
 * The allocated image buffer has to be freed by using
 * av_freep(&pointers[0]).
 *
 * @param align the value to use for buffer size alignment
 * @return the size in bytes required for the image buffer, a negative
 * error code in case of failure
 */
int av_image_alloc(uint8_t *pointers[4], int linesizes[4],
                   int w, int h, enum AVPixelFormat pix_fmt, int align);

/**
 * Copy image plane from src to dst.
 * That is, copy "height" number of lines of "bytewidth" bytes each.
 * The first byte of each successive line is separated by *_linesize
 * bytes.
 *
 * bytewidth must be contained by both absolute values of dst_linesize
 * and src_linesize, otherwise the function behavior is undefined.
 *
 * @param dst_linesize linesize for the image plane in dst
 * @param src_linesize linesize for the image plane in src
 */
void av_image_copy_plane(uint8_t       *dst, int dst_linesize,
                         const uint8_t *src, int src_linesize,
                         int bytewidth, int height);

/**
 * Copy image in src_data to dst_data.
 *
 * @param dst_linesizes linesizes for the image in dst_data
 * @param src_linesizes linesizes for the image in src_data
 */
void av_image_copy(uint8_t *dst_data[4], int dst_linesizes[4],
                   const uint8_t *src_data[4], const int src_linesizes[4],
                   enum AVPixelFormat pix_fmt, int width, int height);

/**
 * Setup the data pointers and linesizes based on the specified image
 * parameters and the provided array.
 *
 * The fields of the given image are filled in by using the src
 * address which points to the image data buffer. Depending on the
 * specified pixel format, one or multiple image data pointers and
 * line sizes will be set.  If a planar format is specified, several
 * pointers will be set pointing to the different picture planes and
 * the line sizes of the different planes will be stored in the
 * lines_sizes array. Call with src == NULL to get the required
 * size for the src buffer.
 *
 * To allocate the buffer and fill in the dst_data and dst_linesize in
 * one call, use av_image_alloc().
 *
 * @param dst_data      data pointers to be filled in
 * @param dst_linesize  linesizes for the image in dst_data to be filled in
 * @param src           buffer which will contain or contains the actual image data, can be NULL
 * @param pix_fmt       the pixel format of the image
 * @param width         the width of the image in pixels
 * @param height        the height of the image in pixels
 * @param align         the value used in src for linesize alignment
 * @return the size in bytes required for src, a negative error code
 * in case of failure
 */
int av_image_fill_arrays(uint8_t *dst_data[4], int dst_linesize[4],
                         const uint8_t *src,
                         enum AVPixelFormat pix_fmt, int width, int height, int align);

/**
 * Return the size in bytes of the amount of data required to store an
 * image with the given parameters.
 *
 * @param pix_fmt  the pixel format of the image
 * @param width    the width of the image in pixels
 * @param height   the height of the image in pixels
 * @param align    the assumed linesize alignment
 * @return the buffer size in bytes, a negative error code in case of failure
 */
int av_image_get_buffer_size(enum AVPixelFormat pix_fmt, int width, int height, int align);

/**
 * Copy image data from an image into a buffer.
 *
 * av_image_get_buffer_size() can be used to compute the required size
 * for the buffer to fill.
 *
 * @param dst           a buffer into which picture data will be copied
 * @param dst_size      the size in bytes of dst
 * @param src_data      pointers containing the source image data
 * @param src_linesize  linesizes for the image in src_data
 * @param pix_fmt       the pixel format of the source image
 * @param width         the width of the source image in pixels
 * @param height        the height of the source image in pixels
 * @param align         the assumed linesize alignment for dst
 * @return the number of bytes written to dst, or a negative value
 * (error code) on error
 */
int av_image_copy_to_buffer(uint8_t *dst, int dst_size,
                            const uint8_t * const src_data[4], const int src_linesize[4],
                            enum AVPixelFormat pix_fmt, int width, int height, int align);

/**
 * Check if the given dimension of an image is valid, meaning that all
 * bytes of the image can be addressed with a signed int.
 *
 * @param w the width of the picture
 * @param h the height of the picture
 * @param log_offset the offset to sum to the log level for logging with log_ctx
 * @param log_ctx the parent logging context, it may be NULL
 * @return >= 0 if valid, a negative error code otherwise
 */
int av_image_check_size(unsigned int w, unsigned int h, int log_offset, void *log_ctx);

/**
 * Check if the given dimension of an image is valid, meaning that all
 * bytes of a plane of an image with the specified pix_fmt can be addressed
 * with a signed int.
 *
 * @param w the width of the picture
 * @param h the height of the picture
 * @param max_pixels the maximum number of pixels the user wants to accept
 * @param pix_fmt the pixel format, can be AV_PIX_FMT_NONE if unknown.
 * @param log_offset the offset to sum to the log level for logging with log_ctx
 * @param log_ctx the parent logging context, it may be NULL
 * @return >= 0 if valid, a negative error code otherwise
 */
int av_image_check_size2(unsigned int w, unsigned int h, int64_t max_pixels, enum AVPixelFormat pix_fmt, int log_offset, void *log_ctx);

2.成员函数

const char *av_get_colorspace_name(enum AVColorSpace val);

AVFrame *av_frame_alloc(void);

void av_frame_free(AVFrame **frame);

int av_frame_ref(AVFrame *dst, const AVFrame *src);

AVFrame *av_frame_clone(const AVFrame *src);

void av_frame_move_ref(AVFrame *dst, AVFrame *src);

int av_frame_get_buffer(AVFrame *frame, int align);

int av_frame_is_writable(AVFrame *frame);

int av_frame_make_writable(AVFrame *frame);

int av_frame_copy(AVFrame *dst, const AVFrame *src);

int av_frame_copy_props(AVFrame *dst, const AVFrame *src);

AVFrameSideData *av_frame_new_side_data(AVFrame *frame,
                                        enum AVFrameSideDataType type,
                                        size_t size);

AVFrameSideData *av_frame_new_side_data_from_buf(AVFrame *frame,
                                                 enum AVFrameSideDataType type,
                                                 AVBufferRef *buf);

void av_frame_remove_side_data(AVFrame *frame, enum AVFrameSideDataType type);

const char *av_frame_side_data_name(enum AVFrameSideDataType type);

上面函数核心就是操作AVFrame。

AVFrame Host内存的获取 av_frame_get_buffer

ffmpeg提供了一个比较特殊的接口

/**
 * Allocate new buffer(s) for audio or video data.
 *
 * The following fields must be set on frame before calling this function:
 * - format (pixel format for video, sample format for audio)
 * - width and height for video
 * - nb_samples and channel_layout for audio
 *
 * This function will fill AVFrame.data and AVFrame.buf arrays and, if
 * necessary, allocate and fill AVFrame.extended_data and AVFrame.extended_buf.
 * For planar formats, one buffer will be allocated for each plane.
 *
 * @warning: if frame already has been allocated, calling this function will
 *           leak memory. In addition, undefined behavior can occur in certain
 *           cases.
 *
 * @param frame frame in which to store the new buffers.
 * @param align Required buffer size alignment. If equal to 0, alignment will be
 *              chosen automatically for the current CPU. It is highly
 *              recommended to pass 0 here unless you know what you are doing.
 *
 * @return 0 on success, a negative AVERROR on error.
 */
int av_frame_get_buffer(AVFrame *frame, int align)

这个函数是用来从ffmpeg内存池中为frame中获取buf，从上面的注释可以看到，如果想要获取buf，必须要设置

 /*
 * The following fields must be set on frame before calling this function:
 * - format (pixel format for video, sample format for audio)
 * - width and height for video
 * - nb_samples and channel_layout for audio
 * /

查看详细的函数代码：

int av_frame_get_buffer(AVFrame *frame, int align)
{
    if (frame->format < 0)
        return AVERROR(EINVAL);

    if (frame->width > 0 && frame->height > 0)
        return get_video_buffer(frame, align);
    else if (frame->nb_samples > 0 && (frame->channel_layout || frame->channels > 0))
        return get_audio_buffer(frame, align);

    return AVERROR(EINVAL);
}

详细的调用流程如下：

AVFrame device内存获取av_hwframe_get_buffer（）

既然有host的内存获取，那么就有device内存的获取

/**
 * Allocate a new frame attached to the given AVHWFramesContext.
 *
 * @param hwframe_ctx a reference to an AVHWFramesContext
 * @param frame an empty (freshly allocated or unreffed) frame to be filled with
 *              newly allocated buffers.
 * @param flags currently unused, should be set to zero
 * @return 0 on success, a negative AVERROR code on failure
 */
int av_hwframe_get_buffer(AVBufferRef *hwframe_ctx, AVFrame *frame, int flags);

其调用流程如下，注意分支代码，这里是继承了父类后，通过mmap去map到父类的内存上去。