On 06/16/2017 04:47 PM, Mike Christie wrote: > The problem with the code currently is that tcmu will send a request to > userspace but not wait for a reply, so it would always return success to > lio. > > This is a problem for dev addition, because userspace could fail, but > the kernel would have a se_device that could get mapped to a lun and > reported to initiators. The requesting app would also get a success status > and it would continue setup. > > For removal this causes a problem where userspace could have done a close > after tcmu does the uio unregister device call. When the uio code is fixed > to not use the parent, then this will cause a crash in uio_release. > > To fix these issues, this patchset made over Nicks for-next branch, > makes the tcmu netlink operations synchronous. > > Patch 1: Fixes for reconfigure added recently. These will be needed in patch5. > Patch 2 - 4: Add target layer support to be able to look up a se_device by id. > Patch 5: Add sync netlink support to tcmu. > Patch 6 - 8: Convert the rest of the target code to use the idr based > id/lookup/iter code. > Patch 9 - 10: These are clean ups to the code that can be done with patches > 1 - 8 in place. They move the rcu call from each backend to lio core. > I forgot to add the tcmu-runner patches for this are at https://github.com/mikechristie/tcmu-runner/tree/sync-nl -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
