I've attached [1] to the bugzilla -- two additional bugfixes I've
found during testing.  One is an access to an initialized variable,
the other sets the correct login state after all sessions are closed.

Thanks,
Kent

Signed-off-by: Kent Yoder <[email protected]>

[1] 
https://sourceforge.net/tracker/download.php?group_id=128009&atid=710344&file_id=409160&aid=3196229

On Mon, Apr 11, 2011 at 2:58 PM, Kent Yoder <[email protected]> wrote:
> Hi,
>
>  This patch changes the way opencryptoki handles sessions by moving
> from using a reference to memory to using a handle which references a
> binary tree node.  This keeps opencryptoki from crashing when an app
> uses an invalid session handle.  testcases/driver/sess_bench reports
> no performance regression using this patch. Due to its size, the patch
> is not inline, but available here [1].
>
> - Uses a binary tree implemented in usr/lib/pkcs11/common/btree.c
> - One tree at the API level and another for each STDLL, replacing a
> linked list at the API level
> - The Session_Struct_t list is totally gone now, replaced by sess_btree
> in the API_Proc_Struct_t. The "value" pointer in the API-level btree
> holds a ref to a ST_SESSION_T (a struct with just a handle and slot ID) and
> the "value" pointer in the STDLL btree is a ref to a SESSION struct.
> - The C_CloseAllSessions was tricky and required me to iterate through
> all the nodes of the tree, checking if the slot ID is matches and
> removing the node if so. For this you'll see a bt_for_each_node()
> function, which uses a callback to check if the slot matches and if so,
> close the session. Calling C_CloseSession from the callback would be
> nicer, but we'd hit a deadlock on SessListMutex.
> - There are a couple of minor clean ups in here too:
>  -- Pass ST_SESSION_T * through various APIs instead of the full struct itself
>  -- Pass a SESSION struct down into the TPM STDLL instead of a ST_SESSION_T
>  -- No more SESS_SET macro, which was ugly :)
>
> Tested on the TPM and software tokens on i686 and the software token on s390x.
>
> Note that this patch doesn't address passing in CK_OBJECT_HANDLE's,
> which also use references to memory.  Another patch for that will
> follow.
>
> Thanks,
> Kent
>
>  usr/include/pkcs11/apictl.h                |   27 +--
>  usr/include/pkcs11/local_types.h           |   31 ++
>  usr/include/pkcs11/stdll.h                 |  112 ++++----
>  usr/lib/pkcs11/aep_stdll/Makefile.am       |    2 +-
>  usr/lib/pkcs11/api/Makefile.am             |    3 +-
>  usr/lib/pkcs11/api/api_interface.c         |  483 +++++++++++----------------
>  usr/lib/pkcs11/api/apiproto.h              |    9 +-
>  usr/lib/pkcs11/api/apiutil.c               |  167 +++++-----
>  usr/lib/pkcs11/bcom_stdll/Makefile.am      |    2 +-
>  usr/lib/pkcs11/cca_stdll/Makefile.am       |    1 +
>  usr/lib/pkcs11/cca_stdll/globals.c         |    2 +-
>  usr/lib/pkcs11/cca_stdll/h_extern.h        |    6 +-
>  usr/lib/pkcs11/cca_stdll/new_host.c        |  281 +++++++---------
>  usr/lib/pkcs11/common/btree.c              |  318 ++++++++++++++++++
>  usr/lib/pkcs11/common/globals.c            |    2 +-
>  usr/lib/pkcs11/common/h_extern.h           |    6 +-
>  usr/lib/pkcs11/common/new_host.c           |  276 +++++++---------
>  usr/lib/pkcs11/common/sess_mgr.c           |  241 ++++++++-------
>  usr/lib/pkcs11/cr_stdll/Makefile.am        |    2 +-
>  usr/lib/pkcs11/ica_s390_stdll/Makefile.am  |    2 +-
>  usr/lib/pkcs11/ica_stdll/Makefile.am       |    2 +-
>  usr/lib/pkcs11/soft_stdll/Makefile.am      |    1 +
>  usr/lib/pkcs11/tpm_stdll/Makefile.am       |    1 +
>  usr/lib/pkcs11/tpm_stdll/globals.c         |    2 +-
>  usr/lib/pkcs11/tpm_stdll/h_extern.h        |    6 +-
>  usr/lib/pkcs11/tpm_stdll/new_host.c        |  275 +++++++---------
>  usr/lib/pkcs11/tpm_stdll/tok_spec_struct.h |    2 +-
>  usr/lib/pkcs11/tpm_stdll/tok_specific.h    |    2 +-
>  usr/lib/pkcs11/tpm_stdll/tpm_specific.c    |    3 +-
>  29 files changed, 1200 insertions(+), 1067 deletions(-)
>
> [1] 
> https://sourceforge.net/tracker/download.php?group_id=128009&atid=710344&file_id=408193&aid=3196229
>

------------------------------------------------------------------------------
Benefiting from Server Virtualization: Beyond Initial Workload 
Consolidation -- Increasing the use of server virtualization is a top
priority.Virtualization can reduce costs, simplify management, and improve 
application availability and disaster protection. Learn more about boosting 
the value of server virtualization. http://p.sf.net/sfu/vmware-sfdev2dev
_______________________________________________
Opencryptoki-tech mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opencryptoki-tech

Reply via email to