name: libaom/scale-allocation-by-element-sizeof description: When allocating dynamic buffers with aom_malloc, aom_memalign, or malloc, multiply element counts by sizeof and pass a single byte-size expression rubric: safety severity: blocking

Rule. When allocating dynamic memory with aom_malloc, aom_memalign, malloc, or realloc, pass a single total byte size computed by multiplying the element count by the element size (e.g., count * sizeof(*ptr) or count * sizeof(Type)). Do not pass an unscaled element count or pass count and size as separate arguments (use aom_calloc(count, size) when allocating with separate count and size parameters).

Why. aom_malloc, aom_memalign, and malloc accept a single argument representing total size in bytes rather than element counts. Allocating without multiplying by sizeof creates an undersized heap buffer, causing out-of-bounds memory corruption. Passing count and element size as separate arguments to aom_malloc causes a signature mismatch.

Examples

Bad:

// Incorrect: Allocating an unscaled element count creates an undersized buffer.
buf2[0] = (float *)aom_malloc(newallocsize);

// Incorrect: Passing separate count and size arguments to a single-argument allocator.
CHECK_MEM_ERROR(cm, cpi->consec_zero_mv,
                aom_malloc(current_size, sizeof(*cpi->consec_zero_mv)));

Good:

// Correct: Scale count by sizeof element to allocate the required bytes.
buf2[0] = (float *)aom_malloc(newallocsize * sizeof(*buf2[0]));

// Correct: Pass a single multiplied byte size expression (or use aom_calloc for separate args).
CHECK_MEM_ERROR(cm, cpi->consec_zero_mv,
                aom_malloc(current_size * sizeof(*cpi->consec_zero_mv)));