system/NetworkManager
system/NetworkManager
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/NetworkManager/NetworkManager synced 2024-07-21 10:14:41 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

glib-aux/trivial: reword code comments about bad random fallback in "nm-random-utils.c"

Browse source
This commit is contained in:
Thomas Haller 2023-02-22 13:49:57 +01:00
parent d359a9a1e7
commit da3c9e470e
No known key found for this signature in database
GPG key ID: 29C2366E4DFC5728
1 changed files with 11 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

20
src/libnm-glib-aux/nm-random-utils.c
View file

@ -187,9 +187,9 @@ _bad_random_init_seed(BadRandSeed *seed)
int seed_idx;
GRand *rand;
/* g_rand_new() reads /dev/urandom, but we already noticed that
/* g_rand_new() reads /dev/urandom too, but we already know that
* /dev/urandom fails to give us good randomness (which is why
* we hit the "bad randomness" code path). So this may not be as
* we hit the "bad random" code path). So this may not be as
* good as we wish, but let's hope that it it does something smart
* to give some extra entropy... */
rand = g_rand_new();
@ -231,14 +231,14 @@ _bad_random_bytes(guint8 *buf, gsize n)
* to give us good randomness. Try our best.
*
* Our ability to get entropy for the CPRNG is very limited and thus the overall
* result will not be good randomness. See _bad_random_init_seed().
* result will be bad randomness.
*
* Once we have some seed material, we combine GRand (which is not a cryptographically
* secure PRNG) with some iterative sha256 hashing. It would be nice if we had
* easy access to chacha20, but it's probably more cumbersome to fork those
* implementations than hack a bad CPRNG by using sha256 hashing. After all, this
* is fallback code to get *some* randomness. And with the inability to get a good
* seed, the CPRNG is not going to give us truly good randomness. */
* is fallback code to get *some* bad randomness. And with the inability to get a good
* seed, any CPRNG can only give us bad randomness. */
{
static BadRandState gl_state;
@ -277,11 +277,11 @@ _bad_random_bytes(guint8 *buf, gsize n)
nm_utils_checksum_get_digest(sum, gl_state.sha_digest.full);
/* gl_state.sha_digest.full and gl_state.rand_vals contain now our
* random values, but they are also the state for the next iteration.
* bad random values, but they are also the state for the next iteration.
* We must not directly expose that state to the caller, so XOR the values.
*
* That means, per iteration we can generate 16 bytes of randomness. That
* is for example required to generate a random UUID. */
* That means, per iteration we can generate 16 bytes of bad randomness. That
* is suitable to initialize a random UUID. */
for (i = 0; i < (int) (NM_UTILS_CHECKSUM_LENGTH_SHA256 / 2); i++) {
nm_assert(n > 0);
buf[0] = gl_state.sha_digest.half_1[i] ^ gl_state.sha_digest.half_2[i]
@ -448,6 +448,8 @@ again_getrandom:
return nm_utils_fd_read_loop_exact(fd, p, n, FALSE);
}
/*****************************************************************************/
guint64
nm_random_u64_range_full(guint64 begin, guint64 end, gboolean crypto_bytes)
{
@ -465,7 +467,7 @@ nm_random_u64_range_full(guint64 begin, guint64 end, gboolean crypto_bytes)
* not crypto_bytes despite being requested). Check errno if you care. */
if (begin >= end) {
/* systemd's random_u64_range(0) is an alias for random_u64_range((uint64_t)-1).
/* systemd's random_u64_range(0) is an alias for nm_random_u64().
* Not for us. It's a caller error to request an element from an empty range. */
return nm_assert_unreachable_val(begin);
}