Jivin David Vrabel lays it down ... > David McCullough wrote: > > Jivin David Vrabel lays it down ... > > > >>The attached patch also fixes up a few other bits and pieces. > > > > At least some of the changes you have made to error returns are incorrect, > > BSD uses positive errno returns, and since these drivers run under the > > crypto framework, they need to return positive errno's. I double check > > them when I merge the patch in. You can check in crypto.c to see how the > > return values are used. > > I only changed ocf-bench.c. At the very least init_module() needs to > return a -ve number on error or modprobe from module-init-tools thinks > it succeeded.
Ah, I must have mis-read it, sorry :-) ocf-bench was a quickie so I could make sure the OCF was driving the HW as hard as possible with little overhead. For the IXP I was getting the same throughput (very slightly slower) than the direct access lib version. Also let me compare the other crypto drivers for straight line speed. > > Has this version helped with your previous issues ? > > It seems to have fixed the "scheduling while atomic" errors, yes. > > I only did some real quick testing with "openssl speed" (on an IXP425) > and only aes128 and 3des seems to show any speed improvements. aes256, > sha1, md5 etc. gave the same performance as software (I'm assuming the > cryptodev engine failed and it defaulted to software). openssl is unlikely to be any different at all. Most of this release was in-kernel operation to get OpenSwan fully running. All the improvements apply to that context as well, Cheers, Davidm -- David McCullough, [EMAIL PROTECTED] Ph:+61 7 34352815 http://www.SnapGear.com Custom Embedded Solutions + Security Fx:+61 7 38913630 http://www.uCdot.org - To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
