Always gather MBEDTLS_ENTROPY_BLOCK_SIZE bytes of entropy #293

Merged

gilles-peskine-arm merged 7 commits from entropy-min into development

2019-11-26 11:43:46 -06:00

gilles-peskine-arm commented

2019-10-09 06:57:50 -05:00

(Migrated from github.com)

Enforce that mbedtls_entropy_func gathers a total of MBEDTLS_ENTROPY_BLOCK_SIZE bytes or more from strong sources. In particular, in the default configuration, on a platform with a single entropy source, mbedtls_entropy_func will now grab 64 bytes from that source (due to the new lower bound of MBEDTLS_ENTROPY_BLOCK_SIZE), whereas it formerly only grabbed 32 bytes (due to the per-source threshold).

Backports: no — this is a security improvement but not a critical fix (32 bytes is enough if the entropy is of good quality) and it could be a problem on platforms where gathering more entropy would take an unacceptable amount of time.

Internal ref: IOTCRYPT-454

Enforce that `mbedtls_entropy_func` gathers a total of `MBEDTLS_ENTROPY_BLOCK_SIZE` bytes or more from strong sources. In particular, in the default configuration, on a platform with a single entropy source, `mbedtls_entropy_func` will now grab 64 bytes from that source (due to the new lower bound of `MBEDTLS_ENTROPY_BLOCK_SIZE`), whereas it formerly only grabbed 32 bytes (due to the per-source threshold). Backports: no — this is a security improvement but not a critical fix (32 bytes is enough if the entropy is of good quality) and it could be a problem on platforms where gathering more entropy would take an unacceptable amount of time. Internal ref: IOTCRYPT-454

k-stachowiak (Migrated from github.com) reviewed

2019-10-29 04:15:36 -05:00

AndrzejKurek (Migrated from github.com) reviewed

2019-11-25 07:49:02 -06:00

library/entropy.c

Is it possible to reach ENTROPY_MAX_LOOP here if ctx->source[i].strong == MBEDTLS_ENTROPY_SOURCE_STRONG is false? There's no way to increase strong_size in such case.

Is it possible to reach `ENTROPY_MAX_LOOP` here if `ctx->source[i].strong == MBEDTLS_ENTROPY_SOURCE_STRONG ` is `false`? There's no way to increase `strong_size` in such case.

gilles-peskine-arm (Migrated from github.com) reviewed

2019-11-25 08:02:11 -06:00

library/entropy.c

Yes, but only if there is no strong source. And in this case the call to entropy_gather_internal would have returned MBEDTLS_ERR_ENTROPY_NO_STRONG_SOURCE.

Yes, but only if there is no strong source. And in this case the call to `entropy_gather_internal` would have returned `MBEDTLS_ERR_ENTROPY_NO_STRONG_SOURCE`.

gilles-peskine-arm commented

2019-11-25 08:04:18 -06:00

(Migrated from github.com)

The CI failures are not just due to unrelated problems that plagued the CI for the past few months. I'm investigating.

AndrzejKurek commented

2019-11-26 06:19:36 -06:00

(Migrated from github.com)

The CI seems to still be failing in strange ways. Are these fails related to the PR?

gilles-peskine-arm commented

2019-11-26 06:22:53 -06:00

(Migrated from github.com)

I think many jobs are failing on -head because it's an old branch that's broken on some boards and incompatible with the current version in mbedtls. Since -merge is passing, the CI is ok.

AndrzejKurek (Migrated from github.com) approved these changes

2019-11-26 07:36:53 -06:00

dgreen-arm (Migrated from github.com) approved these changes

2019-11-26 10:18:19 -06:00