You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, we have a mixture of binding styles in OPTE. We have handspun bindings in e.g., illumos_sys_hdrs and xde::{dls, mac_sys, secpolicy, sys}, while the bindgen'd xde::ip is manually triggered.
While illumos header definitions are fairly stable, I'd rather it be the case that a) all *_sys type definitions are automatically updated for the target system, and b) extra kernel APIs we need to make use of (mac_link_* and friends) are minimal friction to add and that we can be certain of correctness. On point a) in particular, we should really be maximising the use of build.rs-managed _sys crates.
Once complete, these might be good to remove from the repo to help support future kernel/driver work in illumos.
The text was updated successfully, but these errors were encountered:
Bingen can have some pretty major pitfalls. Something worth considering as an alternative is ctf-bindgen. It's still very much at the "prototype an idea and see how it goes" phase, but worth thinking about as a future direction.
Currently, we have a mixture of binding styles in OPTE. We have handspun bindings in e.g.,
illumos_sys_hdrs
andxde::{dls, mac_sys, secpolicy, sys}
, while the bindgen'dxde::ip
is manually triggered.While illumos header definitions are fairly stable, I'd rather it be the case that a) all
*_sys
type definitions are automatically updated for the target system, and b) extra kernel APIs we need to make use of (mac_link_*
and friends) are minimal friction to add and that we can be certain of correctness. On point a) in particular, we should really be maximising the use of build.rs-managed_sys
crates.Once complete, these might be good to remove from the repo to help support future kernel/driver work in illumos.
The text was updated successfully, but these errors were encountered: