Le Mon, Nov 24, 2025 at 09:20:25AM -0800, Paul E. McKenney a écrit : > On Mon, Nov 24, 2025 at 01:04:20PM +0000, Will Deacon wrote: > > On Mon, Nov 10, 2025 at 09:29:43AM -0800, Paul E. McKenney wrote: > > > On Mon, Nov 10, 2025 at 11:24:07AM +0000, Will Deacon wrote: > > > > On Sat, Nov 08, 2025 at 10:38:32AM -0800, Paul E. McKenney wrote: > > > > > On Sat, Nov 08, 2025 at 01:07:45PM +0000, Will Deacon wrote: > > > > > > On Wed, Nov 05, 2025 at 12:32:15PM -0800, Paul E. McKenney wrote: > > > > > > > Some arm64 platforms have slow per-CPU atomic operations, for > > > > > > > example, > > > > > > > the Neoverse V2. This commit therefore moves SRCU-fast from > > > > > > > per-CPU > > > > > > > atomic operations to interrupt-disabled > > > > > > > non-read-modify-write-atomic > > > > > > > atomic_read()/atomic_set() operations. This works because > > > > > > > SRCU-fast-updown is not invoked from read-side primitives, which > > > > > > > means that if srcu_read_unlock_fast() NMI handlers. This means > > > > > > > that > > > > > > > srcu_read_lock_fast_updown() and srcu_read_unlock_fast_updown() > > > > > > > can > > > > > > > exclude themselves and each other > > > > > > > > > > > > > > This reduces the overhead of calls to > > > > > > > srcu_read_lock_fast_updown() and > > > > > > > srcu_read_unlock_fast_updown() from about 100ns to about 12ns on > > > > > > > an ARM > > > > > > > Neoverse V2. Although this is not excellent compared to about > > > > > > > 2ns on x86, > > > > > > > it sure beats 100ns. > > > > > > > > > > > > > > This command was used to measure the overhead: > > > > > > > > > > > > > > tools/testing/selftests/rcutorture/bin/kvm.sh --torture refscale > > > > > > > --allcpus --duration 5 --configs NOPREEMPT --kconfig > > > > > > > "CONFIG_NR_CPUS=64 CONFIG_TASKS_TRACE_RCU=y" --bootargs > > > > > > > "refscale.loops=100000 refscale.guest_os_delay=5 > > > > > > > refscale.nreaders=64 refscale.holdoff=30 > > > > > > > torture.disable_onoff_at_boot > > > > > > > refscale.scale_type=srcu-fast-updown refscale.verbose_batched=8 > > > > > > > torture.verbose_sleep_frequency=8 > > > > > > > torture.verbose_sleep_duration=8 refscale.nruns=100" --trust-make > > > > > > > > > > > > > > Signed-off-by: Paul E. McKenney <[email protected]> > > > > > > > Cc: Catalin Marinas <[email protected]> > > > > > > > Cc: Will Deacon <[email protected]> > > > > > > > Cc: Mark Rutland <[email protected]> > > > > > > > Cc: Mathieu Desnoyers <[email protected]> > > > > > > > Cc: Steven Rostedt <[email protected]> > > > > > > > Cc: Sebastian Andrzej Siewior <[email protected]> > > > > > > > Cc: <[email protected]> > > > > > > > Cc: <[email protected]> > > > > > > > --- > > > > > > > include/linux/srcutree.h | 51 > > > > > > > +++++++++++++++++++++++++++++++++++++--- > > > > > > > 1 file changed, 48 insertions(+), 3 deletions(-) > > > > > > > > > > > > I've queued the per-cpu tweak from Catalin in the arm64 fixes tree > > > > > > [1] > > > > > > for 6.18, so please can you drop this SRCU commit from your tree? > > > > > > > > > > Very good! Adding Frederic on CC since he is doing the pull request > > > > > for the upcoming merge window. > > > > > > > > > > But if this doesn't show up in -rc1, we reserve the right to put it > > > > > back in. > > > > > > > > > > Sorry, couldn't resist! ;-) > > > > > > > > I've merged it as a fix, so hopefully it will show up in v6.18-rc6. > > > > > > Even better, thank you!!! > > > > It landed in Linus' tree here: > > > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/arch/arm64?id=535fdfc5a228524552ee8810c9175e877e127c27 > > Again, thank you, and Breno has started backporting it for use in > our fleet. > > > Please can you drop the SRCU change from -next? It still shows up in > > 20251121. > > This one? > > 11f748499236 ("srcu: Optimize SRCU-fast-updown for arm64") > > if so, Frederic, could you please drop this commit?
Dropped, thanks! (And I'm glad to do so given how error-prone it can be). -- Frederic Weisbecker SUSE Labs
