9a8ecc6a3e
commit 34c86f4c4a7be3b3e35aa48bd18299d4c756064d upstream. The locking in af_alg_release_parent is broken as the BH socket lock can only be taken if there is a code-path to handle the case where the lock is owned by process-context. Instead of adding such handling, we can fix this by changing the ref counts to atomic_t. This patch also modifies the main refcnt to include both normal and nokey sockets. This way we don't have to fudge the nokey ref count when a socket changes from nokey to normal. Credits go to Mauricio Faria de Oliveira who diagnosed this bug and sent a patch for it: https://lore.kernel.org/linux-crypto/20200605161657.535043-1-mfo@canonical.com/ Reported-by: Brian Moyles <bmoyles@netflix.com> Reported-by: Mauricio Faria de Oliveira <mfo@canonical.com> Fixes: 37f96694cf73 ("crypto: af_alg - Use bh_lock_sock in...") Cc: <stable@vger.kernel.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> |
||
---|---|---|
.. | ||
internal | ||
acompress.h | ||
aead.h | ||
aes.h | ||
akcipher.h | ||
algapi.h | ||
authenc.h | ||
b128ops.h | ||
blowfish.h | ||
cast5.h | ||
cast6.h | ||
cast_common.h | ||
cbc.h | ||
chacha20.h | ||
cryptd.h | ||
crypto_wq.h | ||
ctr.h | ||
des.h | ||
dh.h | ||
drbg.h | ||
ecdh.h | ||
engine.h | ||
gcm.h | ||
gf128mul.h | ||
ghash.h | ||
hash.h | ||
hash_info.h | ||
hmac.h | ||
if_alg.h | ||
kpp.h | ||
mcryptd.h | ||
md5.h | ||
morus640_glue.h | ||
morus1280_glue.h | ||
morus_common.h | ||
null.h | ||
padlock.h | ||
pcrypt.h | ||
pkcs7.h | ||
poly1305.h | ||
public_key.h | ||
rng.h | ||
scatterwalk.h | ||
serpent.h | ||
sha.h | ||
sha1_base.h | ||
sha3.h | ||
sha256_base.h | ||
sha512_base.h | ||
skcipher.h | ||
sm3.h | ||
sm3_base.h | ||
sm4.h | ||
twofish.h | ||
xts.h |