On 24 Mar 2010, at 19:59, Andrew Deason wrote:

> 
> Simon or someone else can feel free to correct me... but you need to be
> doing this in the kernel[0], which rules out OpenSSL (as I understand
> it). The only feasible options I remember being discussed were Heimdal's
> hcrypto or something in-tree like Marcus' k5ssl.

As I posted here in October last year, OpenAFS's long term crypto plans centre 
around Heimdal's libhcrypto. libhcrypto supports an EVP style interface, which 
means it will be relatively straightforward to replace it with OpenSSL for 
userspace applications, and should be possible to replace it with native kernel 
crypto operations on platforms where those are available. We don't have any 
interest in maintaining our own crypto library - so we're unlikely to import 
the crypto portions of k5ssl, and we wouldn't want GSoC code targetted against 
this.

> [0] It is technically possible to do the encryption in user-space, if
> you call out to a userspace binary like we do for afsdb lookups. In the
> long run, that approach becomes infeasible...

You definitely wouldn't want to do the block ciphering in userspace - the 
amount of data being thrown across the kernel/userspace boundary would be 
insane. It is possible, however, that we might want to do the public key 
encryption of the per-file-key in a userspace callout.

Cheers,

Simon.

_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel

Reply via email to