Some highlights from NEWS:
** Added OpenSSL 3.0 compatibility.
** Removed OpenSSL 1.0 compatibility.
** Support for FIDO 2.1 "minPinLength" extension.
** Support for COSE_EDDSA, COSE_ES256, and COSE_RS1 attestation.
** Support for TPM 2.0 attestation.
** Support for device timeouts; see fido_dev_set_timeout().
** New API calls:
- es256_pk_from_EVP_PKEY;
- fido_cred_attstmt_len;
- fido_cred_attstmt_ptr;
- fido_cred_pin_minlen;
- fido_cred_set_attstmt;
- fido_cred_set_pin_minlen;
- fido_dev_set_pin_minlen_rpid;
- fido_dev_set_timeout;
- rs256_pk_from_EVP_PKEY.
** Reliability and portability fixes.
** Better handling of HID devices without identification strings; gh#381.
Relnotes: Yes
Sponsored by: The FreeBSD Foundation
Cherry-picked from libfido2 upstream f20a735c0a6f:
iso7816: Avoid storing pointers in a packed structure
On CHERI, and thus Arm's experimental Morello prototype architecture,
pointers are represented as capabilities, which are unforgeable bounded
pointers, providing always-on fine-grained spatial memory safety. The
unforgeability is enforced through the use of tagged memory, with one
validity tag bit per capability-sized-and-aligned word in memory. This
means that storing a pointer to an unaligned location, which is not
guaranteed to work per the C standard, either traps or results in the
capability losing its tag (and thus never being dereferenceable again),
depending on how exactly the store is done (specifically, whether a
capability store or memcpy is used).
However, iso7816 itself does not need to be packed, and doing so likely
causes inefficiencies on existing architectures. The iso7816_header_t
member is packed, and the flexible payload array is a uint8_t (which by
definition has no padding bits and is exactly 8 bits in size and, since
CHAR_BITS must be at least 8, its existence implies that it has the same
representation as unsigned char, and that it has size and alignment 1)
so there will never be any padding inserted between header and payload
(but payload may overlap with padding at the end of the struct due to
how flexible arrays work, which means we need to be careful about our
calculations).
Co-authored-by: pedro martelletto <pedro@yubico.com>