> > From: Jason Gunthorpe <jgg@xxxxxxxxxxxx> > > This make all the context creation uniform by having every driver create a > struct verbs_context using the same startup flow. > > Long ago we had a new and old flow, where the new flow was intended to > support > a stable provider ABI, but now that the ABI is private we do not need that. > > By having every single provider having a verbs_context we can now rely on > things like the priv existing and a sane init ordering where the priv can > exist even before the first command runs. > > Having priv always available is a necessary component to introduce support for > the new ioctl kabi. > > Finally, the last three patches rework the ops setup to use a struct of > function pointers scheme and provides defaults for every function. The > defaults return ENOSYS which fixes a bunch of cases where the user could have > called a function without driver support and got a crash. > > This is github PR: > > https://github.com/linux-rdma/rdma-core/pull/281 > > Jason Gunthorpe (8): > verbs: Always allocate a verbs_context > bnxt_re: Convert from init_context to alloc_context > mlx4: Convert from init_context to alloc_context > mlx5: Convert from init_context to alloc_context > verbs: Remove init_context/uninit_context > verbs: Provide a default implementation for every verbs op > verbs: Convert all providers to use verbs_set_ops > verbs: Remove tests for NULL ops Hey Jason, For what it's worth, I built your ops_fix branch and tested over cxgb4. Working worked, no problems. Tested-by: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx> Steve. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
