Race condition: missing lock protection for dgbl->random_provider read/write (rand_lib.c)

Describe the bug

There is an unsynchronized shared pointer access race condition on dgbl->random_provider inside rand_lib.c.

1. Write site: ossl_rand_check_random_provider_on_load line 1074
   Direct assignment dgbl->random_provider = prov runs without any mutex lock after an if judgment.
2. Read site: RAND_bytes_ex line 424
   Direct bare read of dgbl->random_provider without lock protection.

Concurrent multi-thread access leads to TOCTOU race, torn pointer read/write, memory leak, and SIGSEGV crash.

Code snippet

// Line 1074
if (dgbl->random_provider_name == NULL || dgbl->random_provider != NULL)
    return 1;

dgbl->random_provider = prov;  // No lock wrapped

Start in rand_lib.c at ossl_rand_check_random_provider_on_load around line 1074 and RAND_bytes_ex around line 424, then inspect how related shared provider state is synchronized. Confirm that concurrent reads and writes to dgbl->random_provider are protected consistently and that the race, invalid pointer access, and leak described in the issue no longer occur.