Add lock-free per-CPU primitives to library/cpp/yt/rseq

Introduce AddPerCpu and StorePerCpu over an rseq-sharded per-CPU array.

On the x86-64 Linux fast path the update is committed by a hand-rolled
rseq critical section (non-atomic, migration-safe): addq for the 8-byte
accumulate, movq / movdqu for the 8- or 16-byte store. The kernel
restarts the sequence on preemption or migration, and only one thread
runs on a CPU at a time, so no atomic or lock is needed. Off the fast
path (other arches, no kernel rseq) the operation falls back to an
atomic on the slot indexed by sched_getcpu().

A naturally-aligned 8-byte store is single-copy atomic on x86-64, so it
is never observed torn; the 16-byte store may be, which is acceptable for
a last-writer-wins gauge.
commit_hash:6250f6e9e35cf3895ebafe0b534ec12cca50b03b

1 files changed, 4 insertions, 3 deletions

diff --git a/library/cpp/yt/rseq/rseq.cpp b/library/cpp/yt/rseq/rseq.cpp
index 9ebcb4f45ef..c2098d16d11 100644
--- a/library/cpp/yt/rseq/rseq.cpp
+++ b/library/cpp/yt/rseq/rseq.cpp
@@ -77,9 +77,10 @@ bool OwnsRegistration = false;
 bool RegisterCurrentThread()
 {
     // flags = 0. We pass the shared signature and the standard size so that whoever of
-    // {us, tcmalloc, librseq} runs first registers __rseq_abi and the rest get EBUSY,
-    // which is success for our read-only use (we never run rseq critical sections, so the
-    // signature only ever matters for matching this registration call).
+    // {us, tcmalloc, librseq} runs first registers __rseq_abi and the rest get EBUSY
+    // (success). The signature must be 0x53053053: it is the value emitted before abort_ip
+    // in the per_cpu rseq critical sections (see per_cpu-inl.h), and on a kernel abort the
+    // signature must match the registered one or the kernel delivers SIGSEGV.
     if (::syscall(RseqSyscallNumber, &__rseq_abi, RseqRegistrationSize, 0u, RseqSignature) == 0) {
         return true;
     }