hrm...

our aes code issues CLTS instructions when processing every block.. IIUC (questionable) that was (more or less) because the code predates the kfpu bits...

would that still be necessary if it was updated to use the kfpu bits in the kernel?

jbk as much as I understand, kfpu bits should take care of it.

it'd be nice at some point to improve the performance of the kernel crypto bits

just ENOTIME

jbk: If you're using the kfpu then you won't need to do the kpreempt and clts dance yourself.

It's doing it around every block to allow for pre-emption to happen.

I'm trying to compile OmniOS per the instructions on the illumos.org page for building illumos. Doing ./usr/src/tools/scripts/nightly omnios-illumos-omnios.sh results in an error reading the following: 'fatal: cannot change to 'rev-parse': No such file or directory' \ nightly: Error: Variable CODEMGR_WS not set.'

I want to do the build without the omni tool

you likely need to edit your environment file appropriately (likely starting with either omnios-illumos-omnios.sh or usr/src/tools/env/illumos.sh).

for building illumos-gate you need omnios-illumos-gate, see omnios.org/dev/gate.html

[illumos-gate] 17213 fmd crashes with NULL fabric nodeTab -- Bill Sommerfeld <sommerfeld⊙ho>