Hi everyone, I am looking into possible memory ordering optimizations, and I found that the set_leaf () and set_root_leaf () functions in ip4_mtrie use clib_atomic_cmp_and_swap to update the leaves within the mtrie plys. This leads me to suspect that the mtrie structure is designed to be lock-free, but there are two aspects to ip4_mtrie that I am having trouble understanding:
* clib_atomic_cmp_and_swap is only attempted once in each leaf. There is no code to handle the case when the atomic operation fails. * ASSERTs that the leaves are successfully updated are placed immediately after clib_atomic_cmp_and_swap, implying that the atomic operation will never fail. >From these observations, it appears to me that there can only be one instance >adding to the mtrie at a time. In that case, are the compare-and-swap atomics >necessary? Are there any assumptions about the mtrie structure that I am >missing? Thanks, Jason IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#13540): https://lists.fd.io/g/vpp-dev/message/13540 Mute This Topic: https://lists.fd.io/mt/32562533/21656 Group Owner: vpp-dev+ow...@lists.fd.io Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
