[openssl-users] Apparent Memory Leak When Executing EVP_PKEY_derive (MS vc14, 32 bit, OpenSSL 1.0.2h/m)

Seiberlich, Holger Holger.Seiberlich at bruker.com
Wed Nov 22 12:00:36 UTC 2017


Hi Matt:

Thank you very much for your answer! I did what you told me, and there's no longer an indication of a memory leak.

Sincerely,
Holger 

-----Ursprüngliche Nachricht-----
Von: Matt Caswell [mailto:matt at openssl.org] 
Gesendet: Mittwoch, 22. November 2017 11:50
An: Seiberlich, Holger; openssl-users at openssl.org
Betreff: Re: [openssl-users] Apparent Memory Leak When Executing EVP_PKEY_derive (MS vc14, 32 bit, OpenSSL 1.0.2h/m)

Before you close your application down try calling:

CRYPTO_cleanup_all_ex_data()

See:
https://wiki.openssl.org/index.php/Library_Initialization#Cleanup

The whole dance around init/de-init of the library becomes a whole lot easier in 1.1.0 (it should all happen automatically) - but in 1.0.2 you have to call several functions to properly de-initialise the library.

Matt

On 22/11/17 10:19, Seiberlich, Holger via openssl-users wrote:
> Dear OpenSSL community,
> 
> in a C++ (vc14, OpenSSL 1.0.2h/m) project, I'm trying to negotiate a 
> secret between two parties, using EC crypto. I've used 
> https://wiki.openssl.org/index.php/EVP_Key_Agreement
> as a recipe and wrapped memory allocations into unique_ptrs. The code fulfills its purpose of deriving a shared secret, but memory leak detection with
> 	_CrtSetDbgFlag( _CRTDBG_ALLOC_MEM_DF | _CRTDBG_LEAK_CHECK_DF );
> 	_CrtSetReportMode( _CRT_WARN, _CRTDBG_MODE_FILE );
> 	_CrtSetReportFile( _CRT_WARN, _CRTDBG_FILE_STDOUT ); produces a 
> warning:
> 	Detected memory leaks!
> 	Dumping objects ->
> 	{10316} normal block at 0x00307680, 12 bytes long.
> 	 Data: < v0         > F0 76 30 00 00 00 00 00 0D 00 00 00
> 	{10315} normal block at 0x00351068, 16 bytes long.
> 	 Data: <                > 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> 	{10314} normal block at 0x00350F68, 20 bytes long.
> 	 Data: <    h 5         > 00 00 00 00 68 10 35 00 00 00 00 00 04 00 00 00
> 	{10313} normal block at 0x003076F0, 12 bytes long.
> 	 Data: <    h 5     > 0D 00 00 00 68 0F 35 00 00 00 00 00
> 	{10312} normal block at 0x00310A70, 64 bytes long.
> 	 Data: <                > 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> 	{10311} normal block at 0x00309E98, 96 bytes long.
> 	 Data: <p 1  A    '     > 70 0A 31 00 90 41 20 0F A0 0D 27 0F 08 00 00 00
> 	Object dump complete.
> 
> Let me note two things:
> - The line without which the memory leak disappears is the line triggering the actual secret derivation:
>   if(EVP_PKEY_derive(ctx.get(), skey.get(), &skeylen) <= 0)
> - The size of the leak is independent of how often I run the procedure of generating two sets of keys and then calling DeriveSecret().
> 
> ***
> Therefore, I'm wondering whether I've missed a call to free that I must trigger to deallocate this memory.
> ***
> 
> I have added an allocation hook to check, as an example, when the 96-byte allocation occurs:
> 
> 	[Name]					[Language]
> 	libeay32.dll!default_malloc_ex(unsigned int num, const char * file, int line) Line 79	C
>  	libeay32.dll!CRYPTO_malloc(int num, const char * file, int line) Line 342	C
>  	libeay32.dll!lh_new(unsigned long(*)(const void *) h, int(*)(const void *, const void *) c) Line 120	C
>  	libeay32.dll!ex_data_check() Line 279	C
>  	libeay32.dll!def_get_class(int class_index) Line 316	C
>  	libeay32.dll!int_new_ex_data(int class_index, void * obj, crypto_ex_data_st * ad) Line 421	C
>  	libeay32.dll!CRYPTO_new_ex_data(int class_index, void * obj, crypto_ex_data_st * ad) Line 583	C
>  	libeay32.dll!ECDH_DATA_new_method(engine_st * engine) Line 165	C
>  	libeay32.dll!ecdh_data_new() Line 178	C
>  	libeay32.dll!ecdh_check(ec_key_st * key) Line 215	C
>  	libeay32.dll!ECDH_compute_key(void * out, unsigned int outlen, const ec_point_st * pub_key, ec_key_st * eckey, void *(*)(const void *, unsigned int, void *, unsigned int *) KDF) Line 77	C
>  	libeay32.dll!pkey_ec_derive(evp_pkey_ctx_st * ctx, unsigned char * key, unsigned int * keylen) Line 237	C
>  	libeay32.dll!pkey_ec_kdf_derive(evp_pkey_ctx_st * ctx, unsigned char * key, unsigned int * keylen) Line 252	C
>  	libeay32.dll!EVP_PKEY_derive(evp_pkey_ctx_st * ctx, unsigned char * key, unsigned int * pkeylen) Line 345	C
>  	[.]DeriveSecret() Line 802	C++
> 
> Please see below for how I call EVP_PKEY_derive. Can you tell me what I'm missing? Do you need more info?
> 
> Thank you for looking into this and helping!
> 
> Regards,
> Holger
> 
> 
> template<typename T> using deleted_unique_ptr = 
> std::unique_ptr<T,std::function<void(T*)>>;
>       
> deleted_unique_ptr<EVP_PKEY> m_pOwnKey; deleted_unique_ptr<EVP_PKEY> 
> m_pPeerKey;
> 
> void [...]::DeriveSecret()
> {
>   if (!(m_pOwnKey && m_pOwnKey->pkey.ec))
>   {
>     throw KeyUninitializedError("Secret cannot be derived if own key is missing");
>   }
>   if (!m_pPeerKey)
>   {
>     throw KeyUninitializedError("Secret cannot be derived if peer key is missing");
>   }
>   
>   deleted_unique_ptr<EVP_PKEY_CTX> ctx(EVP_PKEY_CTX_new(m_pOwnKey.get(), NULL), [](EVP_PKEY_CTX* obj)
>   {
>     EVP_PKEY_CTX_free(obj);
>   });
> 
>   size_t skeylen;
>   if(!ctx)
>   {
>     /* Error occurred */
>     throw OpenSslError("Error creating EVP_PKEY_CTX");
>   }
>   if(EVP_PKEY_derive_init(ctx.get()) <= 0)
>   {
>     /* Error */
>     throw OpenSslError("Error in EVP_PKEY_derive_init");
>   }
>   if(EVP_PKEY_derive_set_peer(ctx.get(), m_pPeerKey.get()) <= 0)
>   {
>     /* Error */
>     throw OpenSslError("Error setting peer key in key negotiation");
>   }
>   /* Determine buffer length */
>   if(EVP_PKEY_derive(ctx.get(), NULL, &skeylen) <= 0)
>   {
>     /* Error */
>     throw OpenSslError("Error computing length of shared secret");
>   }
> 
>   if (skeylen != secretSize)
>   {
>     throw OpenSslError("Shared secret has unexpected size");
>   }
> 
>   deleted_unique_ptr<unsigned char> skey((unsigned char*)OPENSSL_malloc(skeylen), [](unsigned char* obj)
>   {
>     OPENSSL_free(obj);
>   });
> 
>   /// @note The derivation of the secret seems to leak memory within OpenSSL
>   if(EVP_PKEY_derive(ctx.get(), skey.get(), &skeylen) <= 0)
>   {
>     /* Error */
>     throw OpenSslError("Error deriving shared secret");
>   }
> 
>   [...]
> }
> 
> 
> 
> 
> Dr. Holger Seiberlich
> Software Developer
> Bruker Optik GmbH
> Rudolf-Plank-Str. 27
> 76275 Ettlingen, Germany
>  
> Phone: +49 7243 504-2626
> Fax: +49 7243 504-42626
> Holger.Seiberlich at bruker.com
> www.bruker.com
> 
> Sitz der Gesellschaft ist Ettlingen, HRB 36 2608 Amtsgericht Mannheim
> Geschäftsführer: Urban Faeh, Marc Beisel, Juergen Srega
> 
> Diese E-Mail und alle Anlagen können Betriebs- oder Geschäftsgeheimnisse, oder sonstige vertrauliche Informationen enthalten. Sollten Sie diese E-Mail irrtümlich erhalten haben, ist Ihnen eine Kenntnisnahme des Inhalts, eine Vervielfältigung oder Weitergabe der E-Mail und aller Anlagen ausdrücklich untersagt. Bitte benachrichtigen Sie den Absender und löschen/vernichten Sie die empfangene E-Mail und alle Anlagen. Vielen Dank. 
> 
> This message and any attachments may contain trade secrets or privileged, undisclosed or otherwise confidential information. If you have received this e-mail in error, you are hereby notified that any review, copying or distribution of it and its attachments is strictly prohibited. Please inform the sender immediately and delete/destroy the original message and any copies. Thank you. 
> 
> 


More information about the openssl-users mailing list