Struct corruption in der_decode_sequence_flexi.c #621

New issue

Closed

opened 2023-06-05 08:14:05 -05:00 by stevemit · 2 comments

stevemit commented

2023-06-05 08:14:05 -05:00

(Migrated from github.com)

Prerequisites

Checked the developer manual
Checked that your issue isn't already filed: https://github.com/issues?utf8=✓&q=repo%3Alibtom%2Flibtomcrypt
Checked that your issue isn't due to the fact that you're using asymmetric cryptography and you forgot linking in and/or setting an MPI provider (usually this causes either random crashes or runtime errors like LTC_ARGCHK 'ltc_mp.name != NULL' failure ...). c.f. Ch. "Math Descriptors" of the developer manual.
Checked that your issue isn't related to TomsFastMath's limitation that PK operations can by default only be done with max. 2048bit keys

Description

Inspecting der_decode_sequence_flexi.c at line 152, when parsing an INTEGER we have

             if ((err = mp_init(&l->data)) != CRYPT_OK) {
                 goto error;
             }

The function mp_init writes to an mp_int structure, but l->data is just a void*. This trashes the fields following data in the ltc_asn1_list struct.

Steps to Reproduce

Code inspection.

Version

1.18.2

Additional Information

A fix would be to point data to an allocated mp_int buffer, and then to free it in der_sequence_free.c.

### Prerequisites * [x] Checked the developer manual * [x] Checked that your issue isn't already filed: https://github.com/issues?utf8=✓&q=repo%3Alibtom%2Flibtomcrypt * [x] Checked that your issue isn't due to the fact that you're using asymmetric cryptography and you forgot linking in and/or setting an MPI provider (usually this causes either random crashes or runtime errors like `LTC_ARGCHK 'ltc_mp.name != NULL' failure ...`). c.f. Ch. "Math Descriptors" of the developer manual. * [x] Checked that your issue isn't related to TomsFastMath's limitation that PK operations can by default only be done with max. 2048bit keys ### Description Inspecting `der_decode_sequence_flexi.c` at line 152, when parsing an INTEGER we have ``` if ((err = mp_init(&l->data)) != CRYPT_OK) { goto error; } ``` The function `mp_init` writes to an `mp_int` structure, but `l->data` is just a `void*`. This trashes the fields following `data` in the `ltc_asn1_list` struct. ### Steps to Reproduce Code inspection. ### Version 1.18.2 ### Additional Information A fix would be to point `data` to an allocated `mp_int` buffer, and then to free it in `der_sequence_free.c`.

sjaeckel commented

2023-06-05 09:09:43 -05:00

(Migrated from github.com)

mp_init() is not the mp_init() you're assuming, but

github.com/libtom/libtomcrypt@fae62af0ab/src/headers/tomcrypt_private.h (L145)

which in turn is MPI provider specific and for LTM declared here:

github.com/libtom/libtomcrypt@fae62af0ab/src/math/ltm_desc.c (L56-L69)

This init() then calls init_mpi() which allocates the struct you're expecting to be corrupted.

github.com/libtom/libtomcrypt@fae62af0ab/src/math/ltm_desc.c (L44-L54)

This references the current state of the develop branch and not the latest release, but the mechanism is still the same.

Please re-open if this doesn't fix this for you.

`mp_init()` is not the `mp_init()` you're assuming, but https://github.com/libtom/libtomcrypt/blob/fae62af0ab16f469c2512ec04575dd60ca018657/src/headers/tomcrypt_private.h#L145 which in turn is MPI provider specific and for LTM declared here: https://github.com/libtom/libtomcrypt/blob/fae62af0ab16f469c2512ec04575dd60ca018657/src/math/ltm_desc.c#L56-L69 This `init()` then calls `init_mpi()` which allocates the struct you're expecting to be corrupted. https://github.com/libtom/libtomcrypt/blob/fae62af0ab16f469c2512ec04575dd60ca018657/src/math/ltm_desc.c#L44-L54 This references the current state of the develop branch and not the latest release, but the mechanism is still the same. Please re-open if this doesn't fix this for you.

stevemit commented

2023-06-05 13:15:43 -05:00

(Migrated from github.com)