Problems with OpenSSL Engine and hashing.

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Problems with OpenSSL Engine and hashing.

Marco GRELLA
Hello everybody,
I have a problem in making our OpenSSL Engine that drives our HW
accelerator work fine for hash (SHA1 in particular).
The problem seems to be related to my "Digest_Copy" or (less likely)
"Digest_Cleanup" implementation (I'll explain this further on).
The Engine works fine for the Cipher algorithms (both just operating on
a file and using s_client/s_server), and works fine for SHA1 when
operating on a file.
Trying to run an s_client / s_server session, I noticed that multiple
context are used and the calls to "Digest_Update" function are mixed, so
I have to maintain coherency in some way.

- 1 -
The easiest (and quickest) way to do this is to buffer the data that I
receive at each call to "Digest_Update", for each context, and ask for a
real hash operation only when I receive the "Digest_Final" for that context.
Doing in this way, everything is ok, both operating on a single file
(here only one context is used) and using s_client / s_server (multiple
contexts).
In this scenario, in the "Digest_Copy" function I make a "memcopy" of
the EVP_MD_CTX->md_data field where our data structure sits, and
moreover I manually allocate and copy the buffer in which I am keeping
the stored data.
In the "Digest_Cleanup" I make a "free" of the buffer in which I keep
the data (it is dynamically allocated) and I set to zero the counters
used to keep track of its size and actual occupation.
In this way, as I said, everything works.

- 2 -
The previous solution is mainly a workaround and has a big disadvantage
if you want to hash large files or amount of data. So I decided to use
the capability of our HW accelerator to save and restore the current
context of the hash block. Here I have some problem, when using
s_client/ s_server. Even if I implement it in the most trivial and
inefficient way (RESTORE/UPDATE/SAVE at *each* call to update) it does
not work.
In this scenario, we have a buffer for the context in our data
structure. I allocate this buffer at the first call to "Digest_Update",
obviously not setting the RESTORE flag for this first call.
I call "free" for this buffer and put it to NULL in the "Digest_Cleanup"
function and, if this buffer is valid (not null) I copy it in the
"Digest_Copy" function, by allocating a proper memory area in the "to"
context.
If I use this solution on a single file, it works, so the SAVE/RESTORE
mechanism works fine (and it has been proved elsewhere). But it fails in
the s_client/s_server test. As the main difference is that here the
"Digest_Copy" and "Digest_Cleanup" function are called, I suppose the
problem is here, even if I do not see many differences with the scenario
number -1-

What am I missing?
Could you help me?
Can someone point me to some useful resource or describe me exactly what
has to be implemented in the Digest_Copy (and Digest_Cleanup) function?
Or do you know another way to avoid the problem of "mixed context"? I mean:
sha1_init(ctx_A)
sha1_update(ctx_A)
sha1_init(ctx_B)
sha1_update(ctx_B)
sha1_update(ctx_A)
....
Thank you very much and best regards,
Marco Grella
______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
User Support Mailing List                    [hidden email]
Automated List Manager                           [hidden email]