On 2017/11/7 1:00, Eric Blake wrote:
> On 11/06/2017 12:21 AM, Longpeng(Mike) wrote: >> Test-crypto-hash calls qcrypto_hash_bytesv/digest/base64 with >> errp=NULL, this will cause a NULL poniter deference if afalg_driver > > s/poniter deference/pointer dereference/ > OK. >> doesn't support requested algos: >> ret = qcrypto_hash_afalg_driver.hash_bytesv(alg, iov, niov, >> result, resultlen, >> errp); >> if (ret == 0) { >> return ret; >> } >> >> error_free(*errp); // <--- here >> >> So we must check 'errp & *errp' before dereference. > > No, if we are going to blindly ignore the error from the hash_bytesv() > call, then we should pass NULL rather than errp. > The 'errp' in this palce is convenient for debug, it can tell us the reason for failure without stepping into afalg_driver's hash_bytesv(). >> >> Signed-off-by: Longpeng(Mike) <longpe...@huawei.com> >> --- >> crypto/hash.c | 4 +++- >> 1 file changed, 3 insertions(+), 1 deletion(-) >> >> diff --git a/crypto/hash.c b/crypto/hash.c >> index ac59c63..c464c78 100644 >> --- a/crypto/hash.c >> +++ b/crypto/hash.c >> @@ -60,7 +60,9 @@ int qcrypto_hash_bytesv(QCryptoHashAlgorithm alg, >> * TODO: >> * Maybe we should treat some afalg errors as fatal >> */ >> - error_free(*errp); >> + if (errp && *errp) { >> + error_free(*errp); >> + } > > Any code that uses 'if (errp && *errp)' is suspicious; I think you need I see, thanks a lot. > a v2 that doesn't try to set errp in the first place, rather than > cleaning it up to ignore it after the fact. > I answered above. How about the following approach? int qcrypto_hash_bytesv(...Error **errp) { #ifdef CONFIG_AF_ALG int ret; Error *err2 = NULL; ret = qcrypto_hash_afalg_driver.hash_bytesv(alg, iov, niov, result, resultlen, &err2); if (ret == 0) { return ret; } error_free(*err2); #endif return qcrypto_hash_lib_driver.hash_bytesv(alg, iov, niov, result, resultlen, errp); } I'll send a V2 if there's no objection, otherwise I'll pick your suggested approach. :) -- Regards, Longpeng(Mike)