Issue #2494
Problems With 64bit Slot IDs With Pkcs11 Plugin
Description
Pkcs#11 defines a (ck_slot_id_t) type as an alias of (unsigned long). Yet the pkcs11 plugin uses the
type (int) in many places instead of a (ck_slot_id_t). Apart from a possible signed/unsigned clash,
(int) might be 32bit while (unsigned long) has 64bit (happens on a 64bit ubuntu 14.04).
As a pkcs11 back end provider, this can only work with the StrongSwan plugin if she restricts herself
to 31bit. This contradicts pkcs#11 specification (AFAIK, unless pointed out otherwise).
Example:
pkcs11_public_key.c(853):
static private_pkcs11_public_key find_key_by_keyid(..., int slot, ...) ..
will later call
C_OpenSession(slot, ..)
which cannot work in general.
Sorry, no patch at the moment due to time restrictions.
History
#1 Updated by Tobias Brunner over 3 years ago
- Status changed from New to Feedback
While I see the theoretical problem, are there even such large slot IDs in practice that this would become an issue?
#2 Updated by Jordan Hrycaj over 3 years ago
yes it is if you run on a system with limited space.
#3 Updated by Jordan Hrycaj over 3 years ago
Sorry wrong thread
Yes it is a problem unless you assume that slots have low integers. In general, an ID might be anything, maybe a hash or a pointer?
#4 Updated by Tobias Brunner over 3 years ago
Yes it is a problem unless you assume that slots have low integers.
Well, low is relative even considering 2^31. But that obviously has been the assumption and experience so far. I've no problem with using the proper type for the slot ID throughout the pkcs11 plugin, however, it will currently not be possible to configure such slot IDs (e.g. via VICI).