Re: [PATCH v9 bpf-next 02/10] libbpf: Support layout section handling in BTF

On 26/02/2026 10:05, [email protected] wrote:
>> diff --git a/tools/lib/bpf/btf.c b/tools/lib/bpf/btf.c
>> index 83fe79ffcb8f..3438cc15bb7a 100644
>> --- a/tools/lib/bpf/btf.c
>> +++ b/tools/lib/bpf/btf.c
> 
> [ ... ]
> 
>> @@ -224,66 +242,95 @@ static void btf_bswap_hdr(struct btf_header *h)
> 
> [ ... ]
> 
>>  static int btf_parse_hdr(struct btf *btf)
>>  {
>> -	struct btf_header *hdr = btf->hdr;
>> +	struct btf_header *hdr = btf->raw_data;
>> +	__u32 hdr_len = hdr->hdr_len;
> 
> [ ... ]
> 
>> +	memcpy(&btf->hdr, hdr, min((size_t)hdr_len, sizeof(struct btf_header)));
> 
> The memcpy here correctly caps the copy at sizeof(struct btf_header),
> but btf->hdr.hdr_len retains the original value from the raw data.
> If the input BTF has a future-extended header where hdr_len exceeds
> sizeof(struct btf_header), only sizeof(struct btf_header) bytes are
> stored, yet hdr_len remembers the larger original value. This
> matters in btf_get_raw_data() below.
> 
> [ ... ]
> 
>> @@ -1601,7 +1664,7 @@ static const void *btf_strs_data(const struct btf *btf)
>>
>>  static void *btf_get_raw_data(const struct btf *btf, __u32 *size, bool swap_endian)
>>  {
>> -	struct btf_header *hdr = btf->hdr;
>> +	const struct btf_header *hdr = &btf->hdr;
>>  	struct btf_type *t;
>>  	void *data, *p;
>>  	__u32 data_sz;
> 
> [ ... ]
> 
>>  	memcpy(p, hdr, hdr->hdr_len);
> 
> When hdr_len > sizeof(struct btf_header), does this read past the
> end of the embedded btf->hdr struct? hdr points to &btf->hdr which
> is sizeof(struct btf_header) bytes, but hdr->hdr_len can be larger
> if the parsed BTF had a future-extended header. The extra bytes
> would come from adjacent struct btf fields (types_data, etc.),
> leaking internal state into the serialized output.
> 
> Previously, btf->hdr pointed directly into raw_data where the full
> header (including any extension bytes) was available. The switch to
> an embedded struct loses those extra bytes, but the serialization
> still tries to write hdr_len bytes.
> 
> One possible fix:
> 
>     memcpy(p, hdr, min((__u32)sizeof(struct btf_header), hdr->hdr_len));
>     if (hdr->hdr_len > sizeof(struct btf_header))
>         memset(p + sizeof(struct btf_header), 0,
>                hdr->hdr_len - sizeof(struct btf_header));
>

yep, this is a good catch, will fix.
 
>>  	if (swap_endian)
>> -		btf_bswap_hdr(p);
>> +		btf_bswap_hdr(p, hdr->hdr_len);
>>  	p += hdr->hdr_len;
> 
> [ ... ]
> 
> 
> ---
> AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
> See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md
> 
> CI run summary: https://github.com/kernel-patches/bpf/actions/runs/22436234361
> 
> AI-authorship-score: low
> AI-authorship-explanation: The patch went through nine review iterations on lore with detailed human reviewer interactions, follows established libbpf patterns, and the subtle header over-read is characteristic of human-authored refactoring.
> issues-found: 1
> issue-severity-score: medium
> issue-severity-explanation: btf_get_raw_data() reads past the embedded btf->hdr struct into adjacent fields when hdr_len exceeds sizeof(struct btf_header), silently leaking internal state into serialized BTF output.