On Sun, 20 Sept 2026 at 20:14, Stephen Hemminger <[email protected]> wrote: > > This patch series completes the removal of rte_smp_XX barrier > functions that was announced over five years ago (20.05). > > In the process several cases of race conditions were found > and fixed. See patches 3, 4, and 17. These are marked with Fixes: > tags and Cc: stable. > > Most of this is mechanical with AI assistance, so strongly > encourage maintainers to take a good look. > > Functional and build run on x86 and ARM native. > Should have no functional change, but as previous cleanups of > atomic have shown performance might be impacted by compiler bugs. > > The lock-free stack patch measured faster than the generic version > it replaces, up to 63% on a 32-core x86 box; the numbers are in > that patch's commit message. > > The RFC removed the barrier test outright. > Konstantin pointed out it still covers the custom x86 seq_cst fence, > so patch 19 now keeps the Peterson lock test and rewrites it around > rte_atomic_thread_fence(rte_memory_order_seq_cst) instead. > Need some careful patch series ordering to keep this all bisect > clean. > > > v2 - address review feedback
I see comments on patch 1. There are two conflicts after merging the rte_atomicXX cleanup (those look simple). The last 3 patches are a single topic. -- David Marchand
