Re: [PATCH 0/4] MR9147: rometadata: Implement CIL assembly/metadata reader + some IMetaDataTables methods.

[Nikolay Sivov (＠nsivov)]

Nikolay Sivov (@nsivov) commented about dlls/rometadata/assembly.c:
> - default:
> -
>    return E_INVALIDARG;
> - }
> -
> - if (!pe_rva_to_offset(sections, num_sections, rva, &offset)) return E_INVALIDARG;
> -
> - cor_hdr = (IMAGE_COR20_HEADER *)(assembly->data + offset);
> - if (cor_hdr->cb != sizeof(IMAGE_COR20_HEADER)) return E_INVALIDARG;
> - if (!(pe_rva_to_offset(sections, num_sections, cor_hdr->MetaData.VirtualAddress, &offset))) return E_INVALIDARG;
> -
> - md_start = assembly->data + offset;
> - md_hdr = (struct metadata_hdr *)md_start;
> - if (md_hdr->signature != METADATA_MAGIC) return E_INVALIDARG;
> -
> - num_streams = *(UINT8 *)(md_start + offsetof(struct metadata_hdr, version[md_hdr->length]) + sizeof(UINT16)); /* Flags */
> - streams_cur = md_start + offsetof(struct metadata_hdr, version[md_hdr->length]) + sizeof(UINT16) * 2; /* Flags + Streams */

I haven't looked at the whole thing, but this section looks as if we were going to assume that file contents are always sensible. And if that's what this is doing, I think it's wrong and reader should always validate offsets and sizes.