diff options
| author | Matt Caswell <matt@openssl.org> | 2023-09-25 16:44:47 +0100 |
|---|---|---|
| committer | Matt Caswell <matt@openssl.org> | 2023-09-27 17:23:04 +0100 |
| commit | be203ea3d3a60a881993d1f7552084996d34ba0a (patch) | |
| tree | 3c080b7221229a5d1b6b2649cba90bc35ed3ced0 /providers | |
| parent | 1a18596149a325a679f8244bac52b6796dfcc48e (diff) | |
Fix a mem leak when the FIPS provider is used in a different thread
We were neglecting to register the main thread to receive thread stop
notifications. This is important if the thread that starts the FIPS
provider is not the same one that is used when OPENSSL_cleanup() is
called.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/21964)
Diffstat (limited to 'providers')
| -rw-r--r-- | providers/fips/fipsprov.c | 9 |
1 files changed, 9 insertions, 0 deletions
diff --git a/providers/fips/fipsprov.c b/providers/fips/fipsprov.c index 607ee11763..7ec409710b 100644 --- a/providers/fips/fipsprov.c +++ b/providers/fips/fipsprov.c @@ -706,6 +706,15 @@ int OSSL_provider_init_int(const OSSL_CORE_HANDLE *handle, fgbl->handle = handle; /* + * We need to register this thread to receive thread lifecycle callbacks. + * This wouldn't matter if the current thread is also the same thread that + * closes the FIPS provider down. But if that happens on a different thread + * then memory leaks could otherwise occur. + */ + if (!ossl_thread_register_fips(libctx)) + goto err; + + /* * We did initial set up of selftest_params in a local copy, because we * could not create fgbl until c_CRYPTO_zalloc was defined in the loop * above. |