On Tue, Sep 28, 2010 at 1:58 PM, Phil Sutter <phil.sut...@viprinet.com> wrote: > Hi! > Please do not apply this patch as is - this is just meant for review, > therefore > I didn't bother removing debug output and dead code. > IMO, a bigger rewrite of the internal routines in cryptodev_main.c is needed > in > order to support async operation in a clean manner: struct crypt_op should be > used for communication with the user only, internally more things need to be > passed around. For zero-copy, this is either the process' corresponding > task_struct and mm_struct OR (when mapping pages at write() time) the set of > user pages and scatterlists. In both cases (copy or not) the given IV needs to > be held separately (the above mentioned copy_from_user problem) and most > importantly some error reporting is needed. > I'm looking forward to hearing your suggestions, hints and flame about my > Kung Fu. :)
I like the implementation and the idea. A question would be why use read()/write() instead say of new ioctls (e.g. CIOCCRYPT_ASYNC)? regards, Nikos _______________________________________________ Cryptodev-linux-devel mailing list Cryptodev-linux-devel@gna.org https://mail.gna.org/listinfo/cryptodev-linux-devel
